Does Money Matter in Education? Bruce D

Does Money Matter in Education? Bruce D. Baker SECOND EDITION

Does Money Matter in Education?

BRUCE D. BAKER THE ALBERT SHANKER INSTITUTE, endowed by the American Federation of Teachers and named in honor of its late president, is a nonprofit, nonpartisan organization dedicated to three themes—children’s education, unions as advocates for quality, and both civic education and freedom of association in the public life of democracies. Its mission is to generate ideas, foster candid exchanges and promote constructive policy proposals related to these issues.

The institute commissions original analyses, organizes seminars, sponsors publications and subsidizes selected projects. Its independent board of directors is composed of educators, business representatives, labor leaders, academics and public policy analysts.

BOARD OF DIRECTORS

Danielle Allen John Jackson Susan B. Neuman Paul E. Almeida Clifford B. Janey Pedro Noguera Anthony Bryk Lorretta Johnson Mary Cathryn Ricker Linda Darling-Hammond Susan Moore Johnson Richard Riley Han Dongfang Richard Kahlenberg William Schmidt Carl Gershman Ted Kirsch Randi Weingarten Sara Goldrick-Rab Stanley S. Litow Deborah L. Wince-Smith Ernest Green Michael Maccoby Sarita Gupta Herb Magidson EXECUTIVE DIRECTOR E. D. Hirsch, Jr. Harold Meyerson Leo E. Casey Sol Hurwitz Daniel Montgomery

This document was written for the Albert Shanker Institute and does not necessarily represent the views of the institute or the members of its board of directors.

Permission is hereby granted to reproduce and distribute copies of the work for nonprofit education purposes, provided that copies are distributed at or below cost, and that the author and copyright notice are included on each copy. Any other distribution of these materials is prohibited without first receiving express written permission. TABLE OF CONTENTS

Executive Summary ...... i

Framing the Question ...... 1

From the Coleman Report to the Production Function ...... 2

Is Aggregate Spending Correlated with Outcome Measures? ...... 3

Do Resources Matter? ...... 5

Do School Finance Reforms Matter? ...... 9

Summing Up the Evidence ...... 15

Concluding Thoughts ...... 19

Appendix: Methods and Measures in Money Matters Questions ...... 21

Endnotes ...... 23

About the Author ...... 39 6 Does Money Matter in Education? Does Money Matter in Education? EXECUTIVE SUMMARY

This second edition policy brief revisits the long and storied literature on whether money matters in providing a quality education. It includes research released since the original brief in 2012 and covers a handful of additional topics. Increasingly, political rhetoric adheres to the unfounded certainty that money doesn’t make a difference in education, and that reduced funding is unlikely to harm educational quality. Such proclamations have even been used to justify large cuts to education budgets over the past few years. These positions, however, have little basis in the empirical research on the relationship between funding and school quality. In the following brief, I discuss major studies on three specific topics: (a) whether how much money schools spend matters; (b) whether specific schooling resources that cost money matter; and (c) whether substantive and sustained state school finance reforms matter. Regarding these three questions, I conclude:

DOES MONEY MATTER? positively associated with student outcomes. Yes. On average, aggregate measures of Again, in some cases, those effects are larger per-pupil spending are positively associated than in others, and there is also variation with improved or higher student outcomes. by student population and other contex- The size of this effect is larger in some stud- tual variables. On the whole, however, the ies than in others, and, in some cases, addi- things that cost money benefit students, and tional funding appears to matter more for there is scarce evidence that there are more some students than for others. Clearly, there cost-effective alternatives. are other factors that may moderate the influence of funding on student outcomes, DO STATE SCHOOL FINANCE such as how that money is spent. In other REFORMS MATTER? words, money must be spent wisely to yield Yes. Sustained improvements to the lev- benefits. But, on balance, in direct tests of el and distribution of funding across local the relationship between financial resources public school districts can lead to improve- and student outcomes, money matters. ments in the level and distribution of student outcomes. While money alone may DO SCHOOLING RESOURCES THAT not be the answer, more equitable and COST MONEY MATTER? adequate allocation of financial inputs to Yes. Schooling resources that cost money, schooling provide a necessary underlying including smaller class sizes, additional sup- condition for improving the equity and ade- ports, early childhood programs and more quacy of outcomes. The available evidence competitive teacher compensation (permit- suggests that appropriate combinations of ting schools and districts to recruit and retain more adequate funding with more account- a higher-quality teacher workforce), are ability for its use may be most promising.

i Does Money While there may in fact be better and more In short, money matters, resources that cost Matter in efficient ways to leverage the education money matter, and a more equitable distri- Education? dollar toward improved student outcomes, bution of school funding can improve out- we do know the following: comes. Policymakers would be well-advised • Many of the ways in which schools to rely on high-quality research to guide the currently spend money do improve critical choices they make regarding school student outcomes. finance. • When schools have more money, they have greater opportunity to spend productively. When they don’t, they can’t. • Arguments that across-the-board budget cuts will not hurt outcomes are completely unfounded.

ii Does Money Matter in Education?

FRAMING THE QUESTION Cuomo opined: “We spend more than any other state in the country. It It is hard to imagine a time in the history of American ain’t about the money. It’s about how you spend it— public education when there has been such a wide- and the results.”6 spread political effort to argue that improving the In an interview with New Jersey Gov. Chris Christie, the quality of schools has little or nothing to do with the Wall Street Journal reported: amount of money spent on public education. “According to Mr. Christie, New Jersey taxpayers are spending $22,000 per student in the Newark school That is, that money simply doesn’t matter. system et less than a third of these students Political certainty regarding the unimportance of money graduate, proving that more money isn’t the answer to 7 for schools and the need for schools to “tighten their better performance.” belts” is frequently grounded in misrepresentations of And in conversations regarding federal education spending total spending growth and test score trends at the national priorities, Virginia Rep. Dave Brat proclaimed: level over the past 30 years. The typical storyline is that “Socrates trained Plato in on a rock and then Plato while spending per pupil has increased dramatically and trained in Aristotle roughly speaking on a rock. So, huge 2 pupil-to-teacher ratios have declined, scores on national funding is not necessary to achieve the greatest minds assessments have stagnated and scores on international and the greatest intellects in history.” [sic]8 assessments have fallen behind the rest of the developed world.3 The conclusion: We’re spending more and more, And so it is: We need only provide a sufficient collection of and not getting results, so it’s clear that money doesn’t rocks to ensure educational adequacy. That is, setting aside make a difference. the modern-day competitive wage required to recruit and retain philosophy instructors of the quality of Socrates and To a large extent, the escalation of rhetoric is a sign of the provide them 1-to-1 student-teacher ratios. times, in terms of both economic and political context. At the outset of the recent economic downturn, U.S. Secretary While political rhetoric is often divorced from empiri- of Education Arne Duncan declared this to be the era of cally rigorous research, the echo chamber regarding the the “new normal,” a period in which budget cutbacks are unimportance of funding for improving school quality has the norm and local public school districts must learn to do amplified, and has migrated to the entirely unsupportable more with less. proposition that funding cuts cause no harm. In other words, the political message has gone several steps beyond At the state level, where the primary responsibility for questioning whether a systematic relationship exists financing public schools lies, this rhetoric has been partic- between funding and school quality—a classic research ularly bold. framing of the issue—to bold assertions that we now know, with certainty, that money doesn’t matter and that the path Florida Gov. Rick Scott, for example, in justifying his cuts to to school improvement can be accomplished despite—or the state’s education budget, remarked: even because of—reductions in spending. “We’re spending a lot of money on education, and when you look at the results, it’s not great.”4 Whether political rhetoric is partly to blame, even as the economy has begun to rebound in most states, state school In his 2011 “State of the State” address, New York Gov. finance systems have become increasingly inequitable, Andrew Cuomo declared: with levels of state support for public schools stagnant at “Not only do we spend too much, but we get too little in best. In a recently published article, I found that during the return. We spend more money on education than any recession, state school finance systems took a substantial state in the nation, and we are number 34 in terms of hit, both in terms of total state and local revenue and in results.”5 terms of equity between districts serving lower- and high- More recently, in reference to a legal challenge brought er-poverty student populations: against New York State by small city school districts, “The recent recession yielded an unprecedented decline

PAGE 1 Does Money in public school funding fairness. Thirty-six states had 1. Does money matter? Are differences in aggregate Matter in a three-year average reduction in current spending school funding associated with differences in short- Education? fairness between 2008-09 and 2010-11, and 32 states and long-term measured outcomes? had a three-year average reduction in state and local revenue fairness over that same time period. Over the 2. Do school resources that cost money matter? entire 19-year period, only 15 states saw an overall Are differences in access to specific schooling decline in spending fairness. In years prior to 2008 programs or resources—where “resources” mean (starting in 1993), only 11 states saw an overall decline the various things that money buys, such as smaller in spending fairness.”9 classes, higher salaries and instructional materials— associated with differences in short- and long-term A more recent report from the Center on Budget and Pol- measured outcomes? icy Priorities revealed that through 2014-15, most state school finance systems had not yet begun to substantively 3. Do school finance reforms matter? Do substantive rebound: and sustained reforms to state school finance “At least 30 states are providing less funding per student systems, including raising the level of funding for the 2014-15 school year than they did before the or redistributing money more equitably, lead to recession hit. Fourteen of these states have cut per- improvements in the level or distribution of student student funding by more than 10 percent. (These outcomes? figures, like all the comparisons in this paper, are in inflation-adjusted dollars and focus on the primary form I discuss only domestic studies, primarily those that focus of state aid to local schools.) on short-term and intermediate-term outcomes, such as achievement (e.g., test scores) and attainment (e.g., grad- Most states are providing more funding per student uation). Furthermore, preference is given to studies that in the new school year than they did a year ago, but appear in peer-reviewed academic journals and books (see funding has generally not increased enough to make endnote 13 for the full selection criteria).13 I also discuss up for cuts in past years. For example, Alabama is the sources of information that have been frequently used increasing school funding by $16 per pupil this year. to cast doubt on whether money is related to education- But that is far less than is needed to offset the state’s al outcomes. Finally, I summarize what we know from $1,144 per-pupil cut over the previous six years.”10 the preponderance of evidence, as derived from rigorous In short, the decline of state school finance systems contin- empirical analysis, as well as what we do not yet know. ues and the rhetoric opposing substantive school finance And in an appendix to this brief, I discuss, in general terms, reform shows little sign of easing. Districts serving the methodological issues around the study of whether money neediest student populations continue to take the hardest matters in education. hit. Yet, concurrently, many states are raising outcome standards for students11 and increasing the consequences on schools and teachers for not achieving those outcome FROM THE COLEMAN standards. Some positive signs include recent structural reforms in California and Pennsylvania, possibly involving REPORT TO THE new revenue, in each case focusing on districts serving high-poverty student populations. But other states that PRODUCTION FUNCTION made substantial cuts during the economic downturn have The saga over whether money matters in American continued to cut (Kansas) or largely freeze (New York) state public education can be traced back to the broader aid, even under the pressure of prior and ongoing judicial question of whether schools matter. That is, whether review and oversight. schools and school quality have any influence on student achievement, educational attainment and The growing political consensus that money doesn’t matter stands in sharp contrast to the substantial body of empir- future earnings. The first national, large-scale quan- ical research that has accumulated over time, but which titative analysis to explore this question was sociolo- gets virtually no attention in our public discourse.12 This gist James Coleman’s widely cited “Equality of Edu- policy brief reviews that literature. Specifically, I review cational Opportunity” report, which came about as three major bodies of evidence, each of which pertains to part of the Civil Rights Act of 1964.14 a specific element of the broad topic of whether money matters in determining the quality of education. These Among other things, the Coleman report explored the three literatures are organized by the following guiding relationship between school resource measures and stu- questions: dent outcomes, finding little relationship between the two.

PAGE 2 Using the (more limited) statistical techniques of the day, evaluating the association between district-level spending Does Money Coleman concluded that, on balance, the strongest cor- and earnings, as opposed to those attempting to identify Matter in relations with student outcome measures were not found a link between school-level resources and earnings, for Education? in schools but rather among factors related to parental which results are murkier.19 income, parental education levels and resources in the home. That said, among school resource measures, Cole- The re-analyses of Coleman’s data, coupled with subse- man did find that teacher characteristics were positively quent credible findings using alternative data sources, associated with student outcomes, and more strongly so for served to discredit the original Coleman report findings minority students compared with white students.15 None- (and more specifically, common interpretations of the theless, the implication drawn by many was that schools report that schools and school quality matter little). It is simply don’t matter. An extension of this implication was now clear that schools matter. that putting more money into schools to try to improve quality was unlikely to matter either.

However, recent re-analyses of the Coleman report data, IS AGGREGATE SPENDING using up-to-date statistical techniques and computing CORRELATED WITH capacity, found that even Coleman’s data indicate that schooling quality has significant effects on student out- OUTCOME MEASURES? comes. In one recent example, Konstantopolous and Bor- After the release of the Coleman report, numerous man (2011) conclude: scholars took advantage of new and richer data sourc- “Our results also indicated that schools play meaningful es. They were largely focused on exploring in greater roles in distributing equality or inequality of depth why schools didn’t seem to matter—the com- educational outcomes to females, minorities, and the mon, and now discredited, interpretation of the Cole- disadvantaged.”16 man report. In 1986, 20 years after Coleman, economist In a related analysis, Borman and Dowling (2010) report: Eric Hanushek published a paper that would arguably “Even after statistically taking into account students’ become the most widely cited source for the claim that family background, a large proportion of the variation money simply doesn’t matter when it comes to improv- among true school means is related to differences ing school quality and student outcomes.20 explained by school characteristics.”17 The paper, a meta-analysis of the large collection of post-Coleman studies, used data from a variety of con- In short, while family background certainly matters most, texts, small and large, in the United States and elsewhere. schools matter as well. Furthermore, there exist substan- Hanushek tallied the findings of those studies. Some found tive differences in school quality that explain a substantial a positive relationship between spending and student portion of the variation in student outcomes. outcomes, while others found no relationship or a negative one. He came to the following conclusion, which was itali- Subsequent studies using alternative data sources explored cized for emphasis in the original publication: the relationship between schooling quality and various “There appears to be no strong or systematic outcomes, including the economic rate of return to school- relationship between school expenditures and student ing (e.g., future earnings). For example, Card and Krueger performance.” (p. 1162)21 (1992) studied the relationship between school quality measures, including pupil-to-teacher ratios and relative In the years that followed, this finding became a mantra for teacher pay, and the rate of return to education for men many politicians and advocates, and it has echoed through born between 1920 and 1949. Card and Krueger found that the halls of state and federal courthouses where school men educated in states with higher-quality schools have funding is deliberated. To this day, it has maintained an a higher return to additional years of schooling. Rates of impressive air of credibility in many circles, although, as return were also higher for individuals from states with discussed below, the analyses behind it were refuted on better-educated teachers.18 numerous occasions by leading scholars in the decade that followed. Furthermore, as also shown below, many of the Similarly, Betts (1996) provided an extensive review of studies originally reviewed by Hanushek, which were pub- the literature that attempts to link measures of schooling lished in the 1960s and 1970s, no longer pass muster meth- quality and adult earnings, including Card and Krueger’s odologically, given advances in data quality, statistical study. Betts explains that, while the overall results of such techniques, and researchers’ understanding of educational studies were mixed, they were generally positive. More production and schooling quality. specifically, he pointed to more positive results for studies

PAGE 3 Does Money In assessing Hanushek’s conclusion, it is important to with variations in student outcomes. Largely confirm- Matter in distinguish between inconsistent findings about the spending the meta-analyses of Greenwald, Hedges and Laine, Education? ing-outcomes relationship on the one hand, and bold dec- Wenglinsky’s analysis found that “per-pupil expenditures larations that money doesn’t matter on the other. Within for instruction and the administration of school districts a developed body of research on almost any topic, there are associated with achievement because both result in is always at least some degree of inconsistency in findings. reduced class size, which raises achievement” (p. 221).24 The key is to adjudicate between studies in terms of their quality and scope, and to assess whether a general con- More recent studies (later 1990s and early 2000s) exam- clusion might be drawn from the preponderance of the ining the relationship between financial resources and high-quality evidence. student outcomes made incremental improvements to production function analyses by (a) adjusting the value Accordingly, the most direct rebuttal to Hanushek’s char- of the education dollar for regional cost variation;25 (b) acterization of the findings of existing research came in a testing alternative “functional forms” of the relationship series of re-analyses by University of Chicago scholars Rob between financial inputs and student outcomes; and (c) Greenwald, Larry Hedges and Richard Laine, who gath- applying other statistical corrections for the measurement ered the studies originally cited by Hanushek in 1986 and of inputs.26 These studies have invariably found a positive, conducted meta-analyses of those that met certain quality statistically significant (though at times small) relationship parameters. They included studies that (a) had appeared in between student achievement gains and financial inputs. a refereed journal or book; (b) used U.S. data; (c) had outcome measures that were some form of academic achieve- They also, however, raised new, important issues about the ment; (d) used data at the district or less-aggregate level; complexities of attempting to identify a direct link between (e) employed a model that controlled for socioeconomic money and student outcomes. These difficulties include characteristics, fit with longitudinal data; and (f) includ- equating the value of the dollar across widely varied geo- ed data that were independent of other data included in graphic and economic contexts, as well as accurately sepa- the universe of studies considered by Hanushek. Notably, rating the role of expenditures from that of students’ family these “quality control measures” pruned a significant share backgrounds, which also play some role in determining local of studies22 used by Hanushek. funding. Most of the studies included in Hanushek’s review suffered from serious data and methodological limitations, Specifically pertaining to aggregate per-pupil spending which have since been addressed in more recent work.27 measures, Greenwald, Hedges and Laine (1996) found that, among statistically significant findings, the vast majority Interest in direct dollar-to-outcomes analysis also stalled of study findings were positive (11:1), and that most of due to the imprecision of data on financial resources avail- the analyses that did not find a statistically discernible able to school sites and children. Most existing financial relationship between spending and outcomes still found a data continue to be reported at the school district level, positive association (p. 368). They concluded: but resources may vary widely across schools within these “Global resource variables such as PPE [per-pupil districts. As a result, questions about whether money mat- expenditures] show strong and consistent relations ters are often restricted to linking district-level funding with achievement. In addition, resource variables that with student-level outcomes, which ignores the manner in attempt to describe the quality of teachers (teacher which district funds are distributed among schools. School- ability, teacher education, and teacher experience) site spending data are increasingly available but have not show very strong relations with student achievement.” generally been the subject of new production function (p. 384) studies. That is, few studies have as yet evaluated the relationship between school-level spending and student-level Digging deeper and exploring the relationship between a outcomes. Instead, researchers have increasingly focused variety of resource and student outcome measures, Gre- on “within school” factors that are thought to influence enwald, Hedges and Laine came to the conclusion that “a student outcomes, including schooling resources, such broad range of resources were positively related to student as class sizes and teacher characteristics, which are often outcomes, with ‘effect sizes’ large enough to suggest that more easily linked in datasets to schools and classrooms.28 moderate increases in spending may be associated with significant increases in achievement” (p. 361).23 This find- To summarize this discussion on whether resources ing stands in sharp contrast to Hanushek’s statement of matter, it is important to recognize that Hanushek’s uncertainty. original conclusion from 1986 was merely a statement of “uncertainty” about whether a consistent Other researchers, including Wenglinsky (1996), went on relationship exists between spending and student to explore with greater precision the measures of finan- outcomes—one that is big enough to be important. cial inputs to schooling that are most strongly associated His conclusion was not that such a relationship does

PAGE 4 not exist. Nor was it a statement that schools with ing more money won’t help.31 In other words, the assertion Does Money fewer resources are better, or that reducing funding is that money spent on the current system doesn’t matter, Matter in can be an effective way to improve schools. but it could if the system was to change. Education?

By the early 2000s, the cloud of uncertainty conjured by Of course, in a sense, this is an argument that money does Hanushek in 1986 had largely lifted in the aftermath of matter. But it also misses the important point about the the various, more rigorous studies that followed, with role of experience and education in determining teachers’ finance scholars using detailed datasets to examine more salaries, and what that means for student outcomes. finely grained relationships between money and student outcomes. While teacher salary schedules may determine pay differentials across teachers within districts, the simple fact The uncertainty has been replaced with an empirically is that where one teaches is also very important in deter- grounded confidence that funding does matter. mining how much he or she makes.32 Arguing over attributes that drive the raises in salary schedules also ignores the bigger question of whether paying teachers more in general might improve the quality of the workforce and, DO RESOURCES MATTER? ultimately, student outcomes. Teacher pay is increasingly Analyzing the relationship between overall spending and uncompetitive with salaries offered by other professions, outcomes is a limited tool. Some things work and others and the “penalty” teachers pay increases the longer they 33 do not—a high-spending state or district that allocates stay on the job. resources to ineffective policies might not show results, A substantial body of literature has accumulated to validate and vice versa. In short, it’s not just how much you spend the conclusion that teachers’ overall wages and relative but how you spend it. Accordingly, both parallel with, wages affect the quality of those who choose to enter the and emergent from, the literature exploring whether teaching profession, and whether they stay once they get aggregate measures of per-pupil spending are positively in. For example, Murnane and Olson (1989) found that sal- associated with student outcomes, there are now numer- aries affect the decision to enter teaching and the duration ous studies of how specific schooling resources affect of the teaching career,34 while Figlio (1997, 2002) and Fer- student outcomes. Typically, these studies have explored guson (1991) concluded that higher salaries are associated 35 measures including (but not limited to): with more qualified teachers. In addition, more recent studies have tackled the specific issues of relative pay noted 1. Teacher salaries; and above. Loeb and Page (2000) showed that: “Once we adjust for labor market factors, we estimate 2. Pupil-to-teacher ratios (class sizes). that raising teacher wages by 10 percent reduces high school dropout rates by 3 percent to 4 percent. Our Both of these resource measures have financial implications. findings suggest that previous studies have failed to Thus, it is natural, when exploring whether money matters, produce robust estimates because they lack adequate to explore whether things that cost money matter.29 controls for non-wage aspects of teaching and market differences in alternative occupational opportunities.”36 In short, while salaries are not the only factor involved, they Teacher quality and wages do affect the quality of the teaching workforce, which in The Coleman report looked at a variety of specific school- turn affects student outcomes. ing resource measures, most notably teacher characteristics, finding positive relationships between these traits Research on the flip side of this issue—evaluating spending and student outcomes. A multitude of studies on the constraints or reductions—reveals the potential harm to relationship between teacher characteristics and student teaching quality that flows from leveling down or reducing outcomes have followed, producing mixed messages as spending. For example, Figlio and Rueben (2001) note: to which characteristics matter most and by how much.30 “Using data from the National Center for Education Statis- Inconsistent findings on the relationship between teacher tics we find that tax limits systematically reduce the average “effectiveness” and how teachers get paid—by experience quality of education majors, as well as new public school 37 and education—added fuel to the “money doesn’t matter” teachers in states that have passed these limits.” fire. Since a large proportion of school spending necessar- Salaries also play a potentially important role in improving ily goes to teacher compensation, and (according to this the equity of student outcomes. While several studies show argument) since we’re not paying teachers in a manner that higher salaries relative to labor market norms can that reflects or incentivizes their productivity, then spend- draw higher-quality candidates into teaching, the evidence

PAGE 5 Does Money also indicates that relative teacher salaries across schools evaluation metrics used for allocating salaries), do we see Matter in and districts may influence the distribution of teaching differences in the production of student outcomes? That is, Education? quality. For example, Ondrich, Pas and Yinger (2008) find can comparable or better outcomes be achieved where the that “teachers in districts with higher salaries relative to summed costs of alternative compensation and producing non-teaching salaries in the same county are less likely to metrics for allocating that compensation are equal to or leave teaching and that a teacher is less likely to change less than current costs? districts when he or she teaches in a district near the top of the teacher salary distribution in that county.”38 Assertions that performance-based pay is necessarily more cost-effective than traditional salary structures also falsely With regard to teacher quality and school racial composi- assume traditional step-and-lane salary schedules to be tion, Hanushek, Kain and Rivkin (2004) note: “A school with monolithic. In practice, salary differentials associated with 10 percent more black students would require about 10 experience and credentials vary widely. Some are com- percent higher salaries in order to neutralize the increased pressed from top to bottom, while others are not, and they probability of leaving.”39 Others, however, point to the lim- may favor experience over credentials or vice versa. Hen- ited capacity of salary differentials to counteract attrition by dricks (2015a, 2015b) explores these issues: compensating for working conditions.40 “Increasing salaries for teachers with 3 or more years of experience differentially retains high-ability teachers, In a perfect world, we could tie teacher pay directly to while higher salaries for teachers with 0-2 years of productivity, but contemporary efforts to do so, including experience differentially retain low-ability teachers. performance bonuses based on student test results, have This likely occurs because higher early-career salaries generally failed to produce concrete results in the United disrupt a positive sorting process that exists among 41 States. Recently published studies of individual and group novice teachers.”47 financial incentives continue to find mixed to null effects,42 although alternative compensation models in some settings have yielded positive results. Dee and Wyckoff (2014) find some evidence that a comprehensive strategy com- That is, one might restructure traditional salary schedules bining teacher evaluation and financial incentives can to achieve gains comparable to or greater than deep- yield marginal improvements to the average rate of student er structural changes to compensation. Hendricks also achievement growth among retained teachers.43 Similarly, finds that changing salary structures may alter recruitment in a study of an Austin, Texas, pay-for-performance (P4P) potential and the recruiting pool: program, Balch and Springer (2015) found that the school “Overall, a 1% increase in base salary for teachers of a district’s “REACH program is associated with positive stu- particular experience level increases the proportion dent test score gains in both math and reading during the of the targeted teachers hired by 0.04-0.08 percentage initial year of implementation. Student test score gains are points. Pay increases have the largest effect on hire maintained in the second year, but we do not find any addi- rates among teachers with 2-3 years of experience tional growth.”44 Sojourner and colleagues (2014) found: and the effect diminishes with experience. I show “Exploiting district variation in participation status and that higher teacher salaries provide a dual benefit of timing, we find evidence that P4P-centered HRM [human retaining and attracting a more effective distribution resource management] reform raises students’ achieve- of teachers. Districts may also improve student ment by 0.03 standard deviations.”45 In a more extreme achievement growth at no cost by reshaping their salary application of financial incentives, characterized as “loss schedules so that they are increasing and concave in aversion,” Fryer and colleagues (2012) study the effect of teacher experience.”48 providing teachers’ bonuses in advance and taking the In the wake of growing literature and policy rhetoric money back if students do not improve sufficiently. They asserting the inefficiency of paying teachers according to find that this approach “yields math test score increases experience and credentials, a handful of new studies have between 0.2 and 0.4 standard deviations. This effect is on surfaced revealing that the gains in student outcomes par with the impact of increasing teacher quality by more resulting from increased teacher experience may extend 46 than one standard deviation.” well beyond the first few years of experience.49 Thus, it would not be entirely inefficient for salaries to continue Still missing in this literature are cost-effectiveness com- scaling upward with increased experience, especially given parisons of the alternatives. That is, if we take the same additional costs of implementing alternative measures on total payroll dollars and allocate those dollars traditionally which to base salaries. Wiswall (2013) finds: across teachers with incremental differences in salaries by experience and credentials held, as opposed to imple- “Using an unrestricted experience model I find that menting those salaries and bonuses by the above alter- for mathematics achievement there are high returns natives (along with paying for the associated costs of the to later career teaching experience, about twice as

PAGE 6 much dispersion in initial teacher quality as previously high ability, Gilpin and Kaganovich assert that “countering Does Money estimated, and a pattern of negative selection where this trend would therefore require an increase in the share Matter in high quality teachers are more likely to exit.”50 of GDP spent on basic education, assuming that the insti- Education? tutional setup of the school system remains unchanged” Papay and Kraft (2015) find: (428). In other words, because talent is becoming more “Consistent with past research, we find that teachers expensive more rapidly in other sectors, more investment, experience rapid productivity improvement early as a share of GDP, may be required merely to maintain edu- in their careers. However, we also find evidence of cation quality. That said, this theoretical exposition, while returns to experience later in the career, indicating that built on much the same research base as I review herein teachers continue to build human capital beyond these and previously, is not fully vetted in the present article. first years.”51 Ladd and Sorensen (2014) find: Rothstein (2012) critiques the presumption that tying “Once we control statistically for the quality of teacher pay directly to measures of performance outcomes individual teachers by the use of teacher fixed effects, would necessarily improve efficiency of money allocated to we find large returns to experience for middle school compensation. He explains: teachers in the form both of higher test scores and “Simulations indicate that labor market interactions improvements in student behavior, with the clearest are important to the evaluation of alternative teacher behavioral effects emerging for reductions in student contracts. Typical bonus policies have very small effects absenteeism. Moreover these returns extend well on selection. Firing policies can have larger effects, if beyond the first few years of teaching.”52 accompanied by substantial salary increases. However, misalignment between productivity and measured Perhaps most importantly, the overall efficiency and effec- performance nearly eliminates the benefits while tiveness of teacher compensation does not depend exclu- preserving most of the costs.”53 sively on the extent to which each dollar allocated to any and every teacher’s salary can be associated precisely with And so it goes—while we have some new evidence that a measurable, marginal gain to the test scores of children alternative compensation methods and evaluation met- linked with that teacher. First, benefits of schooling extend rics may yield some positive results, we do not as of yet beyond short-term achievement gains. Second, teacher have a deeper understanding of the relative cost-effective- compensation exists, and exerts whatever influence it ness of alternatives. Further, we have some evidence that may have, within a complex social and economic system. restructuring compensation—while still based on tradi- Thoughtful expositions considering these complex dynam- tional metrics (experience and credentials)—may have ics are few and far between. Two recent examples, however, positive effects on teacher recruitment and retention. What include a largely theoretical piece, supported by longitudi- we do know in each case is that the overall level of teacher nal descriptive data by Gilpin and Kaganovich (2011), and compensation continues to matter for recruitment of talent a recent National Bureau of Economic Research paper by into the teaching profession, relative to other labor market Rothstein (2012). opportunities. Further, the relative compensation of teachers across settings within labor markets continues to matter. Gilpin and Kaganovich (2011) propose a general equilib- rium model of teacher quantity and quality adopting the To summarize, despite all the uproar about paying teachers premise that teachers’ relative wages (to other sectors) based on experience and education, and its misinterpreta- are critical to maintaining a quality teaching workforce. tions in the context of the “Does money matter?” debate, Additionally, compression of salaries (at the high end) this line of argument misses the point. To whatever degree may reduce retention and recruitment of talented teachers. teacher pay matters in attracting high-quality educators Illustrated in their data, the long-term increase in teacher into the profession and retaining them, it’s less about how quantity has led to lagging wage competitiveness, thus they are paid than how much. Furthermore, the average potentially compromising labor quality. But so too has salaries of the teaching profession, with respect to other growth in wages of competing sectors. They explain that labor market opportunities, can substantively affect the a rise in premium for high ability will outpace that for the quality of entrants to the teaching profession, applicants to average. These increase costs are not offset by technologi- preparation programs and student outcomes. Diminishing cal change, hence an additional downward pressure on the resources for schools can constrain salaries and reduce “real” quality of education inputs. The costs of high quality the quality of the labor supply. Further, salary differen- labor are increasing faster than the resources required to tials between schools and districts might help to recruit pay the premiums that simply maintain quality, and there or retain teachers in high-need settings. In other words, exist no viable technological substitutes to offset those resources used for teacher quality matter. increases. As such, we can expect the real quality of educational inputs to decline. Because of the rise in premium for

PAGE 7 Does Money Class size and teacher quantity Researchers continue to revisit data from the Tennessee Matter in STAR study, which in more recent years has permitted Class size is often characterized as a particularly expensive Education? researchers to explore long-term outcomes of those stu- use of additional school dollars.54 Reducing class sizes obvi- dents experimentally subjected to smaller class sizes. For ously costs money, since you have to hire additional teach- example, Dynarski, Hyman and Schanzenbach (2013) find: ers, but the question of whether it’s expensive must rely on “Assignment to a small class increases the probability detailed comparisons of alternative uses of the same dollars, of attending college by 2.7 percentage points, with or the effects on student outcomes of those alternative uses. effects more than twice as large among blacks. Among Instead, most arguments against class size reduction fre- students enrolled in the poorest third of schools, quently proceed by noting that there are significant costs the effect is 7.3 percentage points. Smaller classes to adding more teachers and classrooms (which is, again, increase the likelihood of earning a college degree an unsurprising revelation),55 followed by a (often vague) by 1.6 percentage points and shift students towards statement as to the differences between the most and least high-earning fields such as STEM (science, technology, “effective” teachers (as measured by their effects on test engineering and mathematics), business and scores). The problem here is that one cannot compare the economics. We find that test score effects at the time of cost-effectiveness of class size reduction with “improving the experiment are an excellent predictor of long-term 59 teacher quality,” which is an outcome, not a concrete policy improvements in postsecondary outcomes.” with measurable costs and benefits. Admittedly, there remain some naysayers on whether class size reduction yields cost-effective benefits in terms What we do know, however, is that ample research indicates of student outcomes. But the findings upon which these that children in smaller classes achieve better outcomes, counterarguments are based often lack the weight of large- both academic and otherwise, and that class size reduction scale randomized studies, such as Tennessee’s Project can be an effective strategy for closing racial and socioeco- STAR, relying instead on natural variations in class sizes 56 nomic achievement gaps. For example, Krueger (1999), in across schools.60 a re-analysis of data from the large-scale randomized Ten- nessee Project STAR class size reduction study, concluded: Assertions of excessive cost and inefficiency of class size “The main conclusions are 1) on average, performance reduction often lack rigorous cost-effectiveness analysis. In a on standardized tests increases by four percentile 2011 brief for the Center for American Progress, for example, points the first year students attend small classes; 2) author Matthew Chingos asserted that class size reduction is the test score advantage of students in small classes the “most expensive school reform.”61 But that same report expands by about one percentile point per year in provided no direct cost or cost-effectiveness comparisons subsequent years; 3) teacher aides and measured between class size reduction and other alternatives. A more teacher characteristics have little effect; 4) class size has recent review by Chingos (2013) published as a policy paper a larger effect for minority students and those on free in the Journal of Policy Analysis and Management criticized lunch.”57 class size reduction as broad state-imposed policy, revisiting the costs and potential downsides of statewide class size Among more recent studies on the topic, also re-evaluat- reduction policies implemented in California and Florida.62 ing the Tennessee STAR data, Konstantopolous and Chun Chingos suggests that estimates of long-term earnings of (2009) wrote: students subjected to class size reduction do not justify “We used data from Project STAR and the Lasting the cost,63 but he also acknowledges that sufficient direct Benefits Study to examine the long-term effects of comparisons between spending on class size reduction and small classes on the achievement gap in mathematics, other alternatives are few and far between. reading, and science scores (Stanford Achievement Test). The results consistently indicated that all types Dynarski, Hyman and Schanzenbach (2013) provide the of students benefit more in later grades from being in most direct cost-effectiveness comparison of class size small classes in early grades. These positive effects are reduction policies with other options for which sufficient significant through grade 8. Longer periods in small data on costs and outcome benefits were available: classes produced higher increases in achievement in “A fair conclusion from this analysis is that the effects later grades for all types of students. For certain grades, we ﬁnd in this paper of class size on college enrollment in reading and science, low achievers seem to benefit alone are not particularly large given the costs of the more from being in small classes for longer periods. program. If focused on students in the poorest third It appears that the lasting benefits of the cumulative of schools, then the cost-effectiveness of class size effects of small classes may reduce the achievement reduction is within the range of other interventions. gap in reading and science in some of the later There is no systematic evidence that early interventions grades.”58 pay of more than later ones when the outcome is

PAGE 8 limited to increased college attendance.”64 finance systems. Therefore, states have the greatest control Does Money over whether local public schools have access to sufficient Matter in It’s true that a large body of the literature on the effective- levels of resources, and whether those resources are dis- Education? ness of class size reduction relies on data from a relatively tributed equitably across children and settings. Further- small group of sources, most notably, the Tennessee STAR more, constitutional protections for children’s access to experiment.65 Further, most class size reduction studies adequate and equitable public schooling exist in state con- finding substantial benefits have focused on class size stitutions, not in the U.S. Constitution. Finally, as indicated reduction in early grades (K-3), and most of these pro- at the outset of this brief, it is at the state level where the grams are pilots implemented on a relatively small scale. (A most raucous rhetoric is occurring around these questions comprehensive review of the literature on class size reduc- of whether money matters in education. State legislatures tion is beyond the scope of this brief, but see endnote 66 for and governors can make or break public schooling, and additional resources.)66 they have.69 It’s also true that reducing class size costs more than not Kevin Welner of the University of Colorado and I published reducing class size. But class size reductions, implemented an extensive review on this specific topic, which appears effectively, have positive effects. As such, one can reason- in the November 2011 issue of Teachers College Record. ably infer that using increased resources to reduce class Among other things, we address the research complexities sizes would have positive effects, or that resources matter. of answering questions about the efficacy of state school While it’s certainly plausible that other uses of the same finance reforms. Those complexities can often be reduced money might be equally or even more effective, there is to asking the right questions about (a) whether substantive little evidence to support this. For example, while we are reforms were actually implemented; (b) when they were quite confident that higher teacher salaries lead to increas- implemented and how long they were sustained; and (c) es in the quality of applicants to the teaching profession who was most affected by the reforms. and increases in student outcomes, we do not know wheth- As with other bodies of literature on the effectiveness of er the same money spent toward salary increases would schooling resources, the research on state school finance achieve better or worse outcomes if it were spent toward reforms is a mixed bag in terms of analytic rigor. Second- class size reduction. Indeed, some have raised concerns hand references to dreadful failures following massive that large-scale class size reductions can lead to unintend- infusions of new funding can often be traced to method- ed labor market consequences that offset some of the gains ologically inept, anecdotal tales of desegregation litigation attributable to class size reduction (such as the inability to in Kansas City, Mo., , or to the court-ordered financing of recruit enough fully qualified teachers).67 And many, over urban districts in New Jersey.70 time, have argued the need for a more precise cost-benefit analysis.68 Still, the preponderance of existing evidence In 2009, Eric Hanushek and a consulting defense attorney for suggests that the additional resources expended on class states facing school funding challenges, Alfred Lindseth of size reductions do result in positive effects. Sutherland Asbill & Brennan, produced a book in which one chapter is dedicated to trying to prove that court-ordered school funding reforms in New Jersey, Wyoming, Ken- DO SCHOOL FINANCE tucky and Massachusetts resulted in few or no measurable improvements.71 These conclusions, however, are based on REFORMS MATTER? little more than a series of graphs of student achievement A particularly relevant question for informing the on the National Assessment of Educational Progress in 1992 current “Does money matter?” debate is whether and 2007. The authors show little change in these states’ scores and conclude that the reforms didn’t work. increased and sustained funding provided through state school finance reforms can improve the level The authors assume that, during this period, each of or distribution of student outcomes, including both the four states infused substantial additional funds into long-term outcomes and short-term shifts in aca- public education in response to judicial orders, and that demic achievement. In other words, does the manner these funds were targeted at low-income and minority in which states distribute money matter? And how students.72’73 They also necessarily assume that in all other can we tell? Findings regarding these specific ques- states that serve as a comparison group, similar changes tions might, most directly, inform state legislative did not occur. Yet they validate neither assertion. debates over tax policy and education spending. In contrast, Welner and I review several studies applying Most funding for public education comes from state and more rigorous and appropriate methods for evaluating the local sources and is under the jurisdiction of state school influence of state school finance reforms. Among these

PAGE 9 Does Money analyses is one national study by Card and Payne (2002) the education gap between children from low-income Matter in that evaluates whether changes in spending inequality and nonpoor families. In relation to current spending Education? generally lead to changes in outcome inequality.74 The levels (the average for 2012 was $12,600 per pupil), this authors measure both the extent and the timing of changes would correspond to increasing per-pupil spending in each. These analyses, while imperfect, rise to a level far permanently by roughly $2,863 per student. above those conducted by Hanushek and Lindseth. Card “Specifically, increasing per-pupil spending by 10 and Payne found “evidence that equalization of spending percent in all 12 school-age years increases the levels leads to a narrowing of test score outcomes across probability of high school graduation by 7 percentage family background groups” (p. 49).75 points for all students, by roughly 10 percentage points Additional compelling studies have been published since for low-income children, and by 2.5 percentage points my review in 2011. In 2014 and 2015, Kirabo Jackson, Ruck- for nonpoor children. er Johnson and Claudia Persico (JJP) released a series of “For children from low-income families, increasing per- NBER working papers and articles summarizing their anal- pupil spending by 10 percent in all 12 school-age years yses of a uniquely constructed national data set in which boosts adult hourly wages by $2.07 in 2000 dollars, or they evaluate the long-term effects of selective, substantial 13 percent.”78 infusions of funding to local public school districts, which occurred primarily in the 1970s and 1980s, on high school The findings of this study have been met with some criti- graduation rates and eventual adult income. Virtues of the cism. Specifically, Eric Hanushek has asserted that these JJP analysis include that the analysis provides clearer link- findings of strong, positive longitudinal effects of school ages than many prior studies between the mere presence of finance reforms on student outcomes, running between “school finance reform,” the extent to which school finance 1972 and 2011, are entirely inconsistent with his char- reform substantively changed the distribution of spending acterization of long-term trends in school spending and and other resources across schools and children, and the national average test scores. According to Hanushek, if the outcome effects of those changes. The authors also go effects JJP describe are real, then the massive infusions of beyond the usual short-run connections between changes funding to public education over time would have mitigat- in the level and distribution of funding, and changes in the ed achievement gaps and substantially driven up national level and distribution of test scores, to evaluate changes in averages. Hanushek explains: the level and distribution of educational attainment, high “Their analysis covers schooling experiences for the school completion, adult wages, adult family income and period 1970-2010. Thus, it is useful to connect these the incidence of adult poverty. estimates to actual funding patterns over the period. Between 1970 and 1990, real expenditure per pupil To do so, the authors use data from the Panel Study of increased not by 10 percent but by over 84 percent. By Income Dynamics on “roughly 15,000 PSID sample mem- 2000, this comparison with 1970 topped 100 percent, bers born between 1955 and 1985, who have been followed and it reached almost 150 percent by 2010. No amount into adulthood through 2011.” The authors’ analysis rests of adjustment for special education, LEP, or what have on the assumption that these individuals, and specific you will make these extraordinary increases in school individuals among them, were differentially affected by funding go away. the infusions of resources resulting from school finance reforms that occurred during their years in K-12 schooling. “If a ten percent increase yields the results calculated One methodological shortcoming of this long-term analy- by Jackson, Johnson, and Persico, shouldn’t we have sis is the imperfect connection between the treatment and found all gaps gone (and even reversed) by now due the population that received that treatment.76 The authors to the actual funding increases? And, even with small matched childhood address data to school district bound- effects on the non-poor, shouldn’t we have seen aries to identify whether a child attended a district likely fairly dramatic improvements in overall educational subject to additional funding as a result of court-mandated and labor market outcomes? In reality, in the face of school finance reform. While imperfect, this approach cre- dramatic past increases in school funding, the gaps ates a tighter link between the treatment and the treated in attainment, high school graduation, and family than exists in many prior national, longitudinal and even poverty have remained significant, largely resisting any state-specific school finance analyses.77 major improvement. And, the stagnating labor market performance for broad swaths of the population has Regarding the effects of school finance reforms on long-term captured considerable recent public and scholarly outcomes, the authors summarize their findings as follows: attention.” “Thus, the estimated effect of a 22 percent increase in Perhaps the most illogical assertion of Hanushek is that per-pupil spending throughout all 12 school-age years applying the effect of funding increases estimated by JJP for low-income children is large enough to eliminate to the actual long-term growth in national average per-pu-

PAGE 10 pil spending would lead to the elimination or reversal associated with about 1.4 more school days, a 4 percent Does Money of achievement gaps. As such, since gaps have not been increase in base teacher salaries, and a 5.7 percent Matter in eliminated or reversed, JJP’s estimates must be wrong. reduction in student-teacher ratios. Because class- Education? Neither JJP’s nor Hanushek’s national average spending size reduction has been shown to have larger effects data indicate that all spending increases from 1970 to 2010 for children from disadvantaged backgrounds, this were targeted to all high-poverty districts nationwide. If provides another possible explanation for our overall Hanushek’s average spending increases were driven as results. much by increases in wealthy (low poverty/minority) dis- “While there may be other mechanisms through which tricts as they were by increases in poorer districts, then increased school spending improves student outcomes, gaps would likely remain constant, all else being equal.79 these results suggest that the positive effects are driven, The identification of substantial gains in lifelong outcomes at least in part, by some combination of reductions in for children in districts that did experience increased fund- class size, having more adults per student in schools, ing indicates that greater gains perhaps could have been increases in instructional time, and increases in teacher achieved for children in lower-wealth, higher-poverty com- salaries that may help to attract and retain a more munities, if funding increases had been more systematical- highly qualified teaching workforce.” ly targeted to those communities, nationwide, throughout the time period. In other words, oft-maligned traditional investments in schooling resources occurred as a result of court-imposed Thus, the critical reviewer must ask whether the data, school finance reforms, and those changes in resources methods and analytic approach applied by JJP are suffi- were likely responsible for the resultant long-term gains in ciently more rigorous, and thus provide more compelling student outcomes. Such findings are particularly consis- evidence, than the long-term trend exposition of Hanushek. tent with recent summaries and updated analyses of data The simple answer to that is yes. on class size reduction.

A secondary critique offered by Hanushek is that the fund- Several state-specific longitudinal studies of school finance ing increases evaluated by JJP occurred largely in the 1970s reforms support the JJP findings. Figlio (2004) explains that and early 1980s, when overall spending was much lower, the influence of state school finance reforms on student thus making marginal gains potentially more important outcomes is perhaps better measured within states over than now, when spending has reached and stabilized at time, explaining that national studies of the type attempted inefficiently high levels across the board. Notably, however, by Card and Payne confront problems that include (a) the JJP’s analyses span a longer period than merely the early enormous diversity in the nature of state aid reform plans; 1970s and also span multiple contexts of higher and lower and (b) the paucity of national student performance data.80 spending over the period. While subsequent replications Accordingly, more recent peer-reviewed studies of state of JJP’s findings and further exploration of their data will school finance reforms have applied longitudinal analyses provide additional insights, the current studies provide within specific states. And several such studies provide compelling evidence that school finance reforms can be compelling evidence of the potential positive effects of leveraged to equalize educational and long-term economic school finance reforms. opportunity. Studies of Michigan school finance reforms of the 1990s JJP also address the question of how money is spent, and, in have shown positive effects on student performance in a response to Hanushek, explain that they too concur that both the previously lowest-spending districts81 and previ- how money is spent is important. An important feature of ously lower-performing districts.82 For instance, Roy (2011) the JJP study is that it does explore the resultant shifts in found that Michigan’s school finance reforms of the 1990s specific schooling resources in response to shifts in funding. led to a significant increase among previously low-spend- For the most part, increased spending led to increases in typ- ing districts. Roy, whose analyses measure both whether ical schooling resources, including higher educator salaries, the policy resulted in changes in funding and who was smaller classes, and longer school days and years. JJP explain: affected, found that Michigan’s school finance plan “was “We find that when a district increases per-pupil quite successful in reducing interdistrict spending dispar- school spending by $100 due to reforms, spending ities. There was also a significant positive effect on student on instruction increases by about $70, spending on performance in the lowest-spending districts as measured support services increases by roughly $40, spending in state tests” (abstract).83 on capital increases by about $10, while there are reductions in other kinds of school spending, on Similarly, Papke (2001), also evaluating Michigan school average. finance reforms of the 1990s, found that “increases in spending have nontrivial, statistically significant effects on “We find that a 10 percent increase in school spending is math test pass rates, and the effects are largest for schools

PAGE 11 Does Money with initially poor performance” (p. 821).84 And no results support the conclusion that Act 60 has Matter in contributed to increased dispersion in performance” (p. Education? Most recently, Hyman (2013) also found positive effects 312).90,91 of these Michigan school finance reforms, but the paper On balance, it is safe to say that a sizeable and grow- raised some concerns regarding the distribution of those ing body of rigorous empirical literature validates effects. Hyman found that much of the increase was target- that state school finance reforms can have substan- ed to schools serving fewer low-income children. However, tive, positive effects on student outcomes, including the study did find that students exposed to “$1,000, or 12%, reductions in outcome disparities and increases in more spending per year during grades four through seven overall outcome levels. It is also safe to say that the experienced a 3.9 percentage point increase in the prob- analyses provided by Hanushek and Lindseth (2009) ability of enrolling in college, and a 2.5 percentage point and others92 who have tried to prove that court-or- increase in the probability of earning a degree” (p. 1).85 dered school funding reforms result in few or no A similar peer-reviewed article by Deke (2003) evaluated measurable improvements offer little credible evi- “leveling up” of funding for very low-spending districts in dence, due to significant methodological omissions. Kansas, following a 1992 lower court threat to overturn the In other words, not only does money matter, but funding formula (without formal ruling to that effect). The reforms that determine how money is distributed article found that a 20 percent increase in spending was matter too. associated with a 5 percent increase in the likelihood of students going on to postsecondary education (p. 275).86 Flipping the function: Elsewhere, three studies of Massachusetts school finance Higher outcomes cost more reforms from the 1990s find similar results. The first, a Earlier in this report, I addressed the education produc- non-peer-reviewed report by Downes, Zabel and Ansel tion function literature that seeks to establish a direct (2009) explored, in combination, the influence on student link between resources spent on schools and districts, outcomes of accountability reforms and changes to school and outcomes achieved by students. Production func- spending. They found that “some of the research findings tion studies include studies of how variation in resources show how education reform has been successful in raising across schools and settings is associated with variations the achievement of students in the previously low-spend- in outcomes across those settings, and whether changes ing districts” (p. 5).87 The second study, an NBER working in resources lead to changes in the level or distribution of paper by Guryan (2001), focused more specifically on the outcomes. redistribution of spending resulting from changes to the state school finance formula. Guryan found that “increases The education cost function is the conceptual flip side in per-pupil spending led to significant increases in math, of the education production function. Like production reading, science, and social studies test scores for 4th- and function research, cost function research seeks to identify 8th-grade students. The magnitudes imply that a $1,000 the link between spending variation and outcome varia- increase in per-pupil spending leads to about a third to a tion, cross-sectionally and longitudinally. The goal of the half of a standard-deviation increase in average test scores. education cost function is to discern the levels of spending It is noted that the state aid driving the estimates is targeted associated with efficiently producing specific outcome to under-funded school districts, which may have atypi- levels (the “cost” per se) across varied geographic contexts cal returns to additional expenditures” (p. 1).88 The most and schools serving varied student populations. Most recent of the three, published in 2014 in the Journal of Edu- published studies applying cost function methodology use cation Finance, found that “changes in the state education multiple years of district-level data, within a specific state aid following the education reform resulted in significantly context, and focus on the relationship between cross-dis- higher student performance” (p. 297).89 trict (over time) variations in spending and outcome levels, considering student characteristics, contextual character- Finally, Downes (2004) conducted earlier studies of Ver- istics such as economies of scale, and labor cost variation. mont school finance reforms of the late 1990s (Act 60), Districts are the unit of analysis because they are the gov- noting: erning unit charged with producing outcomes, raising and “All of the evidence cited in this paper supports the receiving the revenues, and allocating the financial and conclusion that Act 60 has dramatically reduced human resources for doing so. Some cost function studies dispersion in education spending and has done this evaluate whether varied expenditures are associated with by weakening the link between spending and property varied levels of outcomes, all else being equal, while other wealth. Further, the regressions presented in this cost function studies evaluate whether varied expenditures paper offer some evidence that student performance are associated with varied growth in outcomes. has become more equal in the post-Act 60 period.

PAGE 12 The existing body of cost function research has produced current outcomes.97 That is, the goal of the cost model is to Does Money the following (in some cases obvious) findings: identify, among existing “outcome producing units” (dis- Matter in tricts or schools), the more (and less) efficient spending Education? 1. The per-pupil costs of achieving higher-outcome levels associated with given outcomes, where those more goals tend to be higher, across the board, than the efficient spending levels associated with any given out- costs of achieving lower-outcome goals, all else come provide a real-world approximation, approaching being equal.93 the minimum costs of achieving those outcomes. 2. The per-pupil costs of achieving any given level CHL’s empirical critique of education cost function research of outcomes are particularly sensitive to student centers on a falsification test, applying findings from a Cal- population characteristics. In particular, as ifornia study by Jennifer Imazeki (2008).98 CHL’s critique concentrated poverty increases, the costs of was published in a non-peer-reviewed special issue of the achieving any given level of outcomes increase Peabody Journal of Education, based on testimony provid- significantly.94 ed in the state of Missouri and funded by the conservative 99 3. The per-pupil costs of achieving any given level Missouri-based Show-Me Institute. The critique asserts of outcomes are sensitive to district structural that if, as it would appear conceptually, the cost function characteristics, most notably, economies of scale.95 is merely the flip side of the production function, then the magnitude of the spending-to-outcomes relationship should be identical between the cost and production func- Researchers have found cost functions of particular value tions. But, in Imazeki’s attempt to reconcile cost and pro- for evaluating the different costs of achieving specific out- duction functions using California data, the results differed come goals across settings and children. In a review of cost dramatically. That is, if one uses a production function analysis methods in education, Downes (2004) explains: to identify the spending associated with certain outcome “Given the econometric advances of the last decade, the levels, and then the cost function, the results differ dra- cost-function approach is the most likely to give accurate matically. CHL use this finding to assert the failure of cost estimates of the within-state variation in the spending functions as a method and, more generally, the uncertainty needed to attain the state’s chosen standard, if the data are of the spending-to-outcomes relationship. available and of a high quality” (p. 9).96 Duncombe and Yinger (2011), however, explain the fallacy This body of literature also has its detractors, including, of this falsification test, in a non-peer-reviewed special most notably, Robert Costrell, Eric Hanushek and Susanna issue of the same journal.100 They explain that while the Loeb (CHL), who, in a 2008 article, assert that cost func- cost and production functions are loosely flip sides of tions are invalid for estimating costs associated with spe- the same equation, they are not exactly such. Production cific outcome levels. They assert that one cannot possibly models are estimated using some outcome measure as the identify the efficient spending level associated with achiev- dependent variable—that which is predicted by the equa- ing any desired outcome level by evaluating the spending tion. In an education production function studying the behavior of existing schools and districts, whose spending effect of spending on outcomes, the dependent variable is is largely inefficient (because, as discussed above, district predicted as a function of (a) a measure of relevant per-pu- expenditures are largely tied up in labor agreements that, pil spending; (b) characteristics of the student population according to these authors, are in no way linked to the served; and (c) contextual factors that might affect the production of student outcomes). If all schools and dis- value of the dollar toward achieving outcomes (economies tricts suffer such inefficiencies, then one cannot possibly of scale, regional wage variation). discern underlying minimum costs by studying those insti- tutions. However, CHL’s argument rests on the assumption Outcomes = that desired outcomes could be achieved while spending f(Spending, Students, Context) substantially less and entirely differently than any existing The cost model starts out similarly, switching the position school or district spends, all else being equal. As discussed of the spending and outcomes measures, and predicting throughout this report, evidence to this effect is sparse to spending levels as a function of outcomes, students and nonexistent. context factors.

Authors of cost function research assert, however, that the Spending = goal of cost modeling is more modest than exact predic- f(Outcomes, Students, Context) tions of minimum cost, and that much can be learned by better understanding the distribution of spending and out- If it was this simple, then one would expect the statistical comes across existing schools and districts, and the varied relationship between outcomes and spending to be the efficiency with which existing schools and districts achieve same from one equation to the next. But there’s an addi-

PAGE 13 Does Money tional piece to the cost function that, in fact, adds import- Do charter schools prove Matter in ant precision to the estimation of the input to outcome money doesn’t matter? Education? relationship. The above equation is a spending function, whereas the cost function attempts to distill “cost” from Some argue that charter schools generally achieve more spending by addressing the share of spending that may be for less or the same funding than traditional public schools “inefficient.” That is: serving similar student populations, thus validating that money doesn’t matter for improving school quality.102 Cost = Spending – Inefficiency, or The core assumption is that charter schooling improves Spending = Cost + Inefficiency efficiency because the flexibility afforded through chartering permits charter schools to engage in more creative That is, some of the variation in spending is variation that teacher compensation strategies and technological substi- does not lead to variations in the outcome measure. While tution, such as trading small class sizes for efficient use of we don’t really know exactly what the inefficiency is (which technology through blended and online learning. Further, dollars are being spent in ways that don’t improve out- efficiency improvement yielded by charter innovations comes), Duncombe and Yinger suggest that we do know creates competitive pressure on traditional public schools some of the indirect predictors of the likelihood that school to improve.103 However, regarding productivity improve- districts spend more than would be needed to minimally ments from technological substitution, a recent review of achieve current outcomes, and that one can include in the charter school literature by Epple, Romano and Zimmer cost model characteristics of districts that explain a portion (2015) characterized online and cyber charter schools of the inefficient spending. This can be done when the in particular as a “failed innovation, delivering markedly spending measure is the dependent variable, as in the cost poorer achievement outcomes than traditional public function, but not when the spending variable is an inde- schools” (p. 55).104 That said, we do not know if these pendent measure, as in the production function.101 markedly poor achievement outcomes were achieved with Spending = f(Outcomes, Students, Context, markedly fewer resources, and thus, a break-even on efficiency. Inefficiency Factors) When inefficiency factors are accounted for in the spending The assertion of large efficiency gains through chartering is function, the relationship between outcomes and spend- often built on poor and/or misestimation of the resources ing more accurately represents a relationship between received and used by charter schools. Specifically, it is outcomes and costs. This relationship would be expected asserted that charter schools generally receive less funding to be different from the relationship between spending and than do traditional public schools and achieve the same or outcomes (without addressing inefficiency) in a typical better outcomes, thus making them more efficient.105 The production function. first assertion, that charter schools receive less funding and spend less, is certainly not uniformly true.106 Baker, In summary, while education cost function research is Libby and Wiley (2015) explain that charter school spend- not designed to test specifically whether and to what ing varies substantially by context and by operator within extent money matters, the sizeable body of cost function context. Some charter operators, in some contexts, spend literature does suggest that achieving higher educational substantially more than both other charter schools and outcomes, all else being equal, costs more than achieving traditional public schools in the same context, while oth- lower educational outcomes. Further, achieving common ers spend much less. The second assertion, that charters educational outcome goals in settings with concentrat- systematically outperform traditional district schools is ed child poverty, children for whom English is a second also suspect, 107 and the specific assertion that those which language and children with disabilities costs more than do spend much less perform similar to or better than tradi- achieving those same outcome goals with less needy stu- tional district schools stands largely untested. dent populations. Cost models provide some insights into how much more money is required in different settings and A handful of studies identify significant positive achieve- with different children to achieve measured outcome goals. ment effects of schools from the Knowledge Is Power Such estimates are of particular interest in this period of Program (KIPP) network, but this same research provides time when more and more states are migrating toward only weak, imprecise measures of the resources available common standards frameworks and common assessments in these schools.108 Baker, Libby and Wiley (2012, 2015) but are still providing their schools and districts with vastly indicate that KIPP schools in New York and Texas tend to different resources. Cost modeling may provide insights spend substantially more than traditional district schools into just how much more funding may be required for all serving similar populations.109 Similarly, Dobbie and Fryer children to have equal opportunity to achieve these com- (2011) declare that high standards and “no excuses” strate- mon outcome goals. gies of select charter school operators are more important than spending differences in producing improved student

PAGE 14 outcomes.110 But spending measures in the study are poor- students from that spending. Others spend less and do Does Money ly documented and incomplete. Baker, Libby and Wiley’s poorly, and still others spend less but do less poorly than Matter in (2012, 2015) review of financial documents and public data, expected (and are thus more efficient). Still, the variations Education? applying model-based comparisons of school-site expen- in the charter school sector, and the variations across tradi- ditures to schools serving similar student populations, tional public schools, may provide insights down the line in reveals that many of the school operators involved in Dob- how to more effectively and efficiently leverage resources. bie and Fryer’s study spent far more than similar district By and large, charter schools that spend more appear to do schools.111 Baker, Libby and Wiley (2012) also explain that so by providing competitive compensation for their teach- much of the additional spending among high-spending ers, offering longer school days and years, and maintaining charter operators is allocated to maintaining smaller class smaller classes, while those that spend less do so by main- sizes, providing longer school days and years, and paying taining inexperienced staff and high turnover. more to teachers, holding experience and education levels constant, for working those additional hours. That is, the investments by charter operators follow traditional wisdom and are not especially innovative. 112 SUMMING UP THE EVIDENCE Perhaps the strongest empirical evidence of charter school efficiency advantages comes from the work of Gronberg, This brings me to a summary of the evidence on Taylor and Jansen (2012) on Texas charter schools.113 The whether money matters in education. Despite the authors find that, generally, Texas “charter schools are relative consistency of empirical findings over time able to produce educational outcomes at lower cost than regarding (a) whether per-pupil spending itself is traditional public schools—probably because they face related to student outcomes; (b) whether spend- fewer regulations—but are not systematically more effi- ing-related resources, such as teacher wages and cient relative to their frontier than are traditional public class sizes, are related to student outcomes; and schools” (p. 302).114 In other words, while the overall cost of (c) whether improving the adequacy and equity of charter schools is lower for comparable output, the varia- school funding can have positive effects on student tions in relative efficiency among Texas charter schools are substantial. Efficiency is neither uniformly nor consistently outcomes, a persistent cloud of doubt hangs over achieved. political deliberations on school funding. Here, I review briefly the sources of that doubt, relative to Related work by these authors reveals that the lower what we do know with some confidence as well as overall expenses are largely a function of lower salaries what we still have yet to figure out about money and and inexperienced staff (Taylor et al., 2011).115 That is, the student outcomes. difference in total staffing cost, and resulting difference in total instructional expense per pupil, was largely due to the reduced experience levels of teachers, resulting in part Main sources of doubt from the fact that many of the schools existed for fewer The primary source of doubt to this day remains the than 10 years (many fewer than five), in addition to high above-mentioned Eric Hanushek finding in 1986 that turnover among teachers in their first few years. That is, “there appears to be no strong or systematic relationship compensation was held lower not because of creative tech- between school expenditures and student performance.” 117 nological substitution or alternative compensation, but because of relative inexperience and high turnover among This single quote, now divorced entirely from the soundly educators. Epple, Romano and Zimmer (2015) suggest that refuted analyses on which it was based, remains a man- these patterns are similar across studies of charter school tra for those wishing to deny that increased funding for 116 teachers. Thus, estimated efficiency gains, where they schools is a viable option for improving school quality. Add do exist, may rely on maintenance of high turnover and to this statement the occasional uninformative and inflam- relatively inexperienced staff, a questionably scalable and matory anecdote regarding urban district spending and sustainable option. student outcomes in places like Kansas City or New Jersey, or the frequently re-created graphs showing spending and Put simply, research on the charter school sector in the achievement over the past few decades, and one has a aggregate tells us little about whether and to what extent rhetorical war against an otherwise overwhelming body of money matters, or if money can be made to matter more empirical evidence.118 or less than it currently does leveraged through traditional investments in public schooling. Some charter schools While research evidence regarding the importance of fund- spend much more than both other charter schools and ing and specific schooling resources for improving student traditional public schools and appear to yield benefits to outcomes has become clearer with time, Hanushek and a

PAGE 15 Does Money handful of peers have become even more entrenched in “Considerable prior research has failed to find a Matter in their views, as reflected in recent public testimony. Rheto- consistent relationship between school spending and Education? ric among detractors has continued to drift from the cloud student performance, making skepticism about such a of doubt to a rock of certainty. That is, certainty that money relationship the conventional wisdom.”123 has little or no role in improving school quality, and that This time, he anchored that claim only to his 2003 piece school finance reforms that infuse additional funds only (by hyperlink to the “prior research” phrase) on the failure lead to greater inefficiency, having little or no effect on of input-based schooling policies,124 choosing to ignore either equity or adequacy of schooling. Notably, Hanushek entirely the considerably larger body of more rigorous work asserts (now and then) that it’s not that money doesn’t I summarize in my 2012 review on the topic. matter at all, but rather that additional money doesn’t matter on top of the already high levels of spending that The extension of these claims that nearly everyone agrees currently exist across all U.S. schools. there’s no clear relationship between spending and student performance is the assertion that there is broad agreement To summarize, the current dogma of Hanushek includes that how money is spent matters far more than how much the following core tenets: money is available. As phrased by Hanushek in the context 1. Because schools already spend so much and do so of New York state school finance litigation: with such great inefficiency, additional funding is “Virtually all analysts now realize that how money is unlikely to lead to improved student outcomes. spent is much more important than how much is spent.”125 2. How money is used matters much more than how much money is spent. As with the prior declarations, this one is made with the exceedingly bold assertion that virtually all analysts agree 3. Differences in the amount of money some schools on this point—without reference to any empirical evidence have than others are inconsequential, since those to that point (a seemingly gaping omission for a decid- with less may simply make smarter spending edly empirical claim about a supposedly empirical truth). decisions. Further, “how money is spent” is constrained by whether sufficient money is there to begin with. While common sense dictates that how money is spent clearly matters, According the recent rhetoric of Hanushek, these princi- thus making this part of the statement widely agreeable, ples are ironclad. In his own words, they are “conventional this does not preclude the relevance of how much money wisdom” on which “virtually all analysts” agree. They is available to spend. are “commonly believed,” “overall truth” and backed by an “enormous amount of scientific analysis,” “substantial Perhaps most disconcerting is that Hanushek has recent- econometric evidence” and “considerable prior research.” ly extended this argument to declare that equity gaps in For example, in the winter of 2015, in the context of school funding, or measures of them, aren’t an important policy funding litigation in New York state, Hanushek opined: concern either. Specifically, Hanushek proclaims: “An enormous amount of scientific analysis has focused “It also underscores how calculations of equity gaps in on how spending and resources of schools relates to spending, of costs needed to achieve equity, or of costs student outcomes. It is now commonly believed that needed to obtain some level of student performance spending on schools is not systematically related to are vacuous, lacking any scientific basis” (p. 4).126 student outcomes.”119 Put differently, what Hanushek is opining by declaring cal- Yet, the enormous amount of scientific analysis to which culations of equity gaps to be vacuous and lacking scientif- Hanushek referred in his expert testimony was primarily ic basis is that it matters not whether one school or district referenced to a 2003 summary of much of his prior work has more resources than another. Regardless of any spend- from the 1980s, work which has been discredited on ing differences, schools and districts can provide equitable numerous occasions,120 including by research produced in education—toward equitable outcome goals. Those with the last 12 years. Similarly, in the same context (Maisto v. substantively fewer resources simply need to be more effi- State of New York), Hanushek proclaims: cient. Since all public schools and districts are presently so “There has been substantial econometric evidence that inefficient, achieving these efficiency gains through more supports this lack of relationship.” creative personnel policies, such as performance-based pay and dismissal of “bad teachers,” is easily attainable. Hanushek again backs his claims with the same short list 121 of dated self-citation. In an even more recent attempt to Of course, even if we assume that creative personnel poli- rebut a new, major study finding positive effects of school cies yield marginal improvements to efficiency, if schools 122 finance reforms, Hanushek (2015) makes the following with varied levels of resources pursued these strategies version of the same claim:

PAGE 16 with comparable efficiency gains, inequities would remain assessment scores appearing relatively flat, much of which Does Money constant. Requiring those with less to simply be more effi- is achieved via the smoke and mirrors of representing Matter in cient with what they have is an inequitable requirement. spending and outcome data on completely different scales, Education? This argument is often linked in popular media and the and via the failure to adjust appropriately for the chang- blogosphere with the popular book and film Moneyball, ing costs and related obligations of the public education which asserts that clever statistical analysis for selecting system and for the changing demography of the tested high-productivity, undervalued players was the basis for population.131 Oversimplified visuals are used to make the the (short-lived) success in 2002 and 2003 of the low-pay- proclamation that student achievement shows “virtually roll Oakland Athletics baseball team. The flaws of this no change,” a statement discredited on closer inspection.132 analogy are too many to explore thoroughly herein, but the Jackson, Johnson and Persico (2015) provide additional biggest flaw is illustrated by the oft-ignored subtitle of the examples of how such facile analyses lead to fallacious book: The Art of Winning an Unfair Game. That is, gaining conclusions.133 a leg up through clever player selection is necessary in baseball because vast wealth and payroll differences across As explained by Baker and Welner (2011),134 Hanushek for teams make baseball an unfair game. The public’s interest years has cited the failures of New Jersey’s school finance in providing equitable and adequate funding for education reforms as the basis for why other states should not is likely greater than ensuring equitable and adequate increase funding to high-poverty schools. In litigation in baseball payrolls. Put more bluntly, the education of pres- Kansas in 2011, Hanushek proclaimed: ent and future generations should not be an unfair game. “The dramatic spending increases called for by the courts (exhibit 34) have had little to no impacts on From judges to scholars, critics of Hanushek have charac- achievement. Compared to the rest of the nation, terized his evidence as “facile,” based on “fuzzy logic,”127 performance in New Jersey has not increased across and “weak and factually tenuous.”128 Two recurring exam- most grades and racial groups (exhibits 35-40). These ples used by Hanushek to illustrate the unimportance results suggest caution in considering the ability of of funding increases for improving outcomes are the courts to improve educational outcomes.”135 “long-term trend” or “time trend” argument and anec- Hanushek reiterated these claims in the context of a even dotal claims of the failures of input-based reforms in more recent New York school funding challenge.136 This New Jersey. Baker and Welner (2011) tackle in depth the is a surprising claim to preserve when one’s own recent fallacies of Hanushek’s New Jersey claims.129 Here, I point (2012), marginally more rigorous analyses of state achieve- to Hanushek’s own, albeit facile, unacknowledged self-de- ment growth rates on national assessments (from 1992 to bunking of his New Jersey claims. But first, I address the 2011)137 find the following: long-term trend claim. “The other seven states that rank among the top-10 Again, from recent testimony in New York state, Hanushek improvers, all of which outpaced the United States as provides the following exposition of the long-term trend a whole, are Massachusetts, Louisiana, South Carolina, assertion: New Jersey, Kentucky, Arkansas, and Virginia.”138 “The overall truth of this disconnect of spending and Further, the same report reveals that New Jersey has seen outcomes is easiest to see by looking at the aggregate particularly strong growth in reducing the number of the data for the United States over the past half century. lowest-performing students (those scoring at the “below Since 1960, pupil‐teacher ratios fell by one‐third, basic” level), especially for eighth-grade math. teachers with master’s degrees over doubled, and median teacher experience grew significantly (Chart To be sure, there are others in academe and policy research 1).4 Since these three factors are the most important that raise questions about the most effective ways to lever- determinants of spending per pupil, it leads to the age school funding to achieve desired outcomes, and do quadrupling of spending between 1960 and 2009 (after so via more rigorous, thoughtful analyses. The most recent adjusting for inflation). At the same time, plotting rigorous and relevant academic research is addressed in scores for math and reading performance of 17‐year‐ the remainder of this brief. There are others who opine in olds on the National Assessment of Educational the public square139 and courtroom140 that school finance Progress (NAEP, or “The Nation’s Report Card”) shows reform—specifically infusing additional funding to districts virtually no change since 1970 (Charts 2 and 3).5”130 serving high-need student populations—is neither the most effective nor the most efficient path toward improving This claim, like many of Hanushek’s, is made with language schooling equity or adequacy. But empirical evidence to of astounding certainty—the “overall truth” as it exists in support claims of more efficient alternatives remains elusive. the mind of Hanushek. This claim is commonly accompanied by graphs showing per-pupil spending going up No rigorous empirical study of which I am aware validates over time, pupil-to-teacher ratios going down and national that increased funding for schools in general, or targeted to

PAGE 17 Does Money specific populations, has led to any substantive, measured characterizations (accompanied by distorted visuals) of Matter in reduction in student outcomes or other “harm.” Arguably, if the long-term trends in student outcomes and school Education? this were the case, it would open new doors to school finance spending. Facile international comparisons are equally litigation against states that choose to increase funding to deceitful, in that they (a) fail to account for differences schools. Twenty years ago, economist Richard Murnane sum- in student populations served and the related scope of marized the issue exceptionally well when he stated: educational and related services provided; and (b) fail to “In my view, it is simply indefensible to use the results appropriately equate educational spending across nations, of quantitative studies of the relationship between including the failure to account for the range of services school resources and student achievement as a basis and operating costs covered under “educational expense” for concluding that additional funds cannot help public in the United States versus other countries (for example, school districts. Equally disturbing is the claim that the public employee health and pension benefits). And anec- removal of funds … typically does no harm” (p. 457).141 dotal assertions of failures resulting from massive infusions of funding are rebutted herein and elsewhere.142 Murnane’s quote is as relevant today as it was then. The sources of doubt on the “Does money matter?” question Finally, while the assertion that “how money is spent is are not credible. important” is certainly valid, one cannot reasonably make the leap to assert that how money is spent is necessarily While there remains much to debate, discuss and empiri- more important than how much money is available. Yes, cally evaluate regarding the returns to each additional dollar how money is spent matters, but if you don’t have it, you spent in schools—and the strategies for improving educa- can’t spend it. It is unhelpful at best for public policy, and tional efficiency, equity and adequacy—we must finally be harmful to the children subjected to those policies, to pre- willing to cast aside the most inane arguments and sources tend without any compelling evidence that somewhere out of evidence on either side of the debate. Specifically, the fol- there exists a far cheaper way to achieve the same or better lowing five contentions no longer have a legitimate place in outcomes (and thus we can cut our way down that more the debate over state school finance policy and whether and efficient path). As so eloquently noted by a three-judge how money matters in K-12 education: panel in Kansas when faced with this question: 1. Vote counts of correlational studies between “Simply, school opportunities do not repeat themselves spending and outcomes, without regard for rigor of and when the opportunity for a formal education the analyses and quality of the data on which they passes, then for most, it is most likely gone. We all know depend; that the struggle for an income very often—too often— overcomes the time needed to prepare intellectually for 2. The long-term trend argument and supporting a better one. graphs that show long-term spending going up and NAEP scores staying flat; “If the position advanced here is the State’s full position, it is experimenting with our children which have no 3. International comparisons asserting, and perhaps recourse from a failure of the experiment.”143 illustrating via scatterplot, that the United States spends more than other developed countries but What do we know? achieves less on international assessments; Based on the studies reviewed in this brief, there are a few things we can say with confidence about the 4. Anecdotal assertions that states such as New Jersey relationship between funding, resources and student and cities such as Kansas City provide proof positive outcomes. that massive infusions of funding have proven ineffective at improving student outcomes; and First, on average, even in large-scale studies across multiple contexts, aggregate measures of per-pupil 5. The assertion that how money is spent is much more spending are positively associated with improved important than how much is available. and/or higher student outcomes. In some studies, the size of this effect is larger than in others, and, in some cases, additional funding appears to matter Vote count tallies without regard for study quality and rigor more for some students than for others. Clearly, are of relative little use for understanding whether money there are other factors that moderate the influence matters in schooling and are of no use for discerning how. of funding on student outcomes, such as how that The long-term trend argument is perhaps the most reiter- money is spent. But, on balance, in direct tests of the ated of all arguments that money doesn’t matter, but it is relationship between financial resources and student built largely on deceptive, oversimplified and largely wrong outcomes, money matters.

PAGE 18 Second, schooling resources that cost money, includ- ally represent an argument that money matters, not the Does Money ing class size reductions and increased teacher com- opposite. Matter in pensation, are positively associated with student Education? outcomes. Again, these effects are larger in some There is also limited evidence about the connection cases and for some populations. On balance, though, between funding and longer-term outcomes. In an era there are ways to spend money that have a solid where educational output and outcomes are increasingly track record of success. Further, while there may measured in terms of short-term changes in students’ per- exist alternative uses of financial resources that yield formance on standardized tests of reading and math, we comparable or better returns in student outcomes, have arguably lost sight of broader and/or intermediate- no clear evidence identifies what these alternatives and long-term outcomes. We need to know more about might be. the relationship between access to resources in preschool, elementary school and secondary school and successful Third, sustained improvements to the level and dis- transitions to and completion of undergraduate education tribution of funding across local public school dis- (and labor market outcomes). We do have a growing body tricts can lead to improvements in the level and dis- of evidence that students’ access to advanced coursework tribution of student outcomes. While money alone in mathematics does have a positive relationship to under- may not be the answer, adequate and equitable graduate success, and that access to a breadth of curricular distributions of financial inputs to schooling provide and co-curricular opportunities increases college access.144 a necessary underlying condition for improving the And we know that such opportunities are inequitably dis- adequacy and equity of outcomes. That is, if the tributed across children.145 This research must expand to money isn’t there, schools and districts simply don’t include a broader array of both inputs and outputs. have a “leverage option” that can support strategies that might improve student outcomes. If the money The primary problem is that state data systems provide lim- is there, they can use it productively; if it’s not, they ited capacity to track students from K-12 systems through can’t. But, even if they have the money, there’s no college and into the workforce. Moreover, while the pre- guarantee that they will use it productively. Evi- cision of financial data is improving in some regards, it dence from Massachusetts, in particular, suggests remains difficult to tie district-level expenditure data to that appropriate combinations of more funding with specific schools, programs and classrooms, limiting the more accountability may be most promising. ability of researchers to explore more closely the relationship between spending patterns, resource allocation choices and student outcomes. Hopefully, states will improve What don’t we know? the quality and scope of their available data in the near future. Indeed, there are many unanswered questions about how money matters and how it can matter most. Specifically, while many talk of more efficient or cost-effective options for spending money, information on these options is sorely CONCLUDING THOUGHTS lacking. Rhetoric abounds regarding current approaches to public schooling—such as spending on class size Given the preponderance of evidence that resources reduction—being the most inefficient or least cost-effec- do matter and that state school finance reforms can tive options. But proposed alternatives, such as restruc- effect changes in student outcomes, it seems some- turing teacher pay around indicators of “effectiveness” what surprising that not only has doubt persisted, rather than seniority or credentials, are not backed by but the rhetoric of doubt seems to have escalated. In solid research and include no serious evaluations of cost. many cases, direct assertions are made that schools Accordingly, they provide no legitimate basis for compar- can do more with less money; that money is not a ing cost-effectiveness. necessary underlying condition for school improvement; and, in the most extreme cases, that cuts to While we do have evidence that increased salaries may funding might actually stimulate improvements that improve the quality of the teacher workforce and student outcomes, we do not have sufficient evidence to determine past funding increases have failed to accomplish. whether the same dollar spent on salaries to “improve To be blunt, money does matter. Schools and districts with teacher quality” by some (often unstated) means would more money clearly have a greater ability to provide high- achieve better or worse outcomes than if that dollar was er-quality, broader and deeper educational opportunities spent on a more proven intervention, such as class size to the children they serve. Furthermore, in the absence of reductions. Moreover, even if there were evidence that money, or in the aftermath of deep cuts to existing funding, some new policy was more cost-effective, this would actu- schools are unable to do many of the things they need to

PAGE 19 Does Money do in order to maintain quality educational opportunities. Matter in Without funding, efficiency tradeoffs (like focusing on Education? teacher quality versus teacher quantity) and innovations (like online learning) being broadly endorsed are suspect. One cannot trade spending money on class size reductions to increase teacher salaries to improve teacher quality if funding is not there for either—if class sizes are already large and teacher salaries noncompetitive. While these are not the conditions faced by all districts, they are faced by many.

It is certainly reasonable to acknowledge that providing more money, by itself, is not a comprehensive solution for improving school quality. Clearly, money can be spent poorly and have limited influence on school quality. On the flip side, money can be spent well and have substantive positive influence. However, money that’s not there can’t do either. The available evidence leaves little doubt: Sufficient financial resources are a necessary underlying condition for providing quality education.

PAGE 20 Appendix: Methods and Measures in Money Matters Questions Does Money Matter in Education? Measuring the inputs Resources that cost money In this appendix, in order to help readers better Differences in school- and district-level instructional understand the methods used in the studies discussed spending often boil down to differences in in this paper, I provide a more detailed primer on quantities of instructional staff and differences in studying the relationship between money and student the characteristics of those staff (most related to outcomes. differences in salaries related to differences in years of experience and degree levels). Quantities of Broadly, studies of the “Does money matter?” genre instructional staff are most often measured in terms seek to determine whether differences or changes in of class sizes or pupil-to-teacher ratios. To the extent access to schooling inputs are associated with or result that having a greater quantity of teachers affects in differences in or changes to student outcomes. Any student outcomes, then so too does having the money such studies must therefore include some measures available to increase the quantity of teachers. of schooling inputs and of student outcomes. In studies that might fall into the “Does money matter?” Teacher experience and degree levels are also often category, input measures include money itself and studied in the context of the “Does money matter?” resources that cost money. debate because, within traditional teacher salary schedules, teachers with more experience and with Money itself more advanced degrees are generally paid higher salaries. To the extent that these characteristics are Per-pupil expenditure is a commonly used measure associated with differences in student outcomes, of the aggregate level of financial resources expenditures on these characteristics may be assumed available in public school districts. The measure to be associated with student outcomes. typically includes all current operating expenditures of school districts—that is, the fiscal year spending One might also look specifically at comprehensive on salaries and benefits for school employees, on school reform models, some of which are noted for classroom supplies and equipment, and on facility their resource intensiveness, such as the Roots and utilities, maintenance and operations—divided by Wings/Success for All model148 and the more recently the number of children served. But this measure touted Apollo 20 program in Houston.149 To the extent is problematic on a number of levels. First, very that these models require greater expenditure than few studies appropriately adjust the value of current levels, and result in better outcomes than per-pupil spending for differences (such as levels current levels, a reasonable argument can be made of labor competition or other costs) across labor that money spent on these reforms matters. Many markets within states.146 Second, some substantive comprehensive reform strategies embed some degree differences in school district offerings that do cost of additional staffing (instructional quantity) with money don’t show up as per-pupil expenditure some degree of professional development (improving variation, such as the addition of prekindergarten instructional quality), and the relative costs of these programs, which adds both spending and students, components may be distilled. often at lower per-pupil spending than occurs in upper grades. It is a substantive addition to the Measuring the outcomes educational program that may, in some cases, reduce Equally pertinent is the measurement of outcomes. average per-pupil spending districtwide. Outcome measures in “Does money matter?” or “Does school quality matter?” studies tend to take Components of per-pupil spending, such three forms: short-term and concurrent academic as “instructional spending” or “administrative achievement measures, mid-term academic attainment spending,” are also occasionally explored for their measures, and long-term economic benefit measures. differential effects (if any) on student outcomes.147 It is often presumed that instructional spending Short-term and concurrent academic achievement differences will be most related to student outcomes measures are the most common in the past two (where instructional spending is often described decades because of the increased availability as “money to the classroom,” consisting of teacher of individual student-level data on academic wages, materials, supplies, equipment and classroom achievement, largely from state data systems support staff). implemented for accountability purposes but also

PAGE 21 Does Money from large national surveys, including the National served and contextual factors of schools and districts in Matter in Education Longitudinal Study of the eighth-grade which those students are served. When framed this way, Education? class of 1988. Typically, when longitudinal data are the statistical models are “production function” models, available on individual students on measures of or models of the production of student outcomes.150 academic achievement, the goal is to determine These studies seek to identify whether there exists a the influence of differential school resources as a statistically significant relationship between spending treatment on gains in student achievement outcomes. measures or other school resource measures and Most commonly, the measured outcomes are for math student outcomes, ideally measured at the individual and language arts. student level and measured in terms of outcome gains. Further, even if statistically significant, it is important Mid-term academic attainment measures include to know how differences in inputs are associated measures of high school graduation rates, transition with differences in outcomes. For example, how many to higher education, persistence in higher education more dollars does it take to improve achievement by a (and completion of specific coursework and credits) specific amount? and time to completion of postsecondary education. These intermediate measures of attainment are Numerous technical issues complicate these less common, perhaps due to the relatively limited analyses, such as problems with fully accounting for availability of detailed individual-level data linking “unobservable” differences in student backgrounds K-12 education system parameters and college or schooling contexts, and difficulties determining attendance patterns of graduates of specific K-12 what the right “shape” of the statistical relationship is schools and districts. between inputs and outcomes (for example, to what extent are there diminishing returns and when do they Long-term economic benefit measures have been the kick in?), each of which may compromise the validity of focus of numerous large-scale economic studies of findings.151 the influence of schooling quality. From an economic perspective, there is great interest in validating that Another type of model, not often discussed as a measurable differences in school quality or investment method for determining whether money matters, is the in schooling can ultimately have measurable effects on education cost function.152 The education cost function both individual wages and the economy as a whole. essentially turns the education production function around in an attempt to determine the costs per pupil Research methods of achieving desired educational outcome levels, given for linking the two the student populations served and contextual factors such as differences in the prices of schooling inputs, A handful of research methods and statistical economies of scale, population sparsity and remoteness. approaches have been used to evaluate the In effect, these studies attempt to determine whether connection between money and schooling resources it costs more to achieve more, and how much it costs and student outcomes. These methods may be broadly to do so, given the average costs of existing practices classified into those that involve studying the “natural of schooling.153 In other words, does money matter? variation” in schooling quality available to individuals, based on where they attend school, and studies that Related studies of existing or historical variation of involve the random assignment of students to receive resources across children have explored the relationship specific reforms, strategies or programs (with fiscal between changes in the distribution or overall level implications). Note that natural variation is a research of funding allocated by states to local public schools euphemism for the vast systemic inequity of the or districts and resulting changes in the level or American public education system. Studies of natural distribution of student outcomes. For example, if a variation may explore differences across schooling state allocates substantially more resources than in contexts or changes in schooling quality over time, the past to low-wealth school districts, do student which are in effect policy-induced variations. outcomes in those districts improve? These are policy- induced variations or changes, but are not experiments. Studies relying on natural variation I refer to these studies as “Do school finance reforms Most studies exploring the relationship between matter?” studies, and they are a particularly relevant existing differences in schooling resources and existing variation on the broader “Does money matter?” differences in student outcomes attempt to estimate question. They are important because state school some form of a statistical model that relates student finance policy is the primary vehicle for changing outcomes to financial or other schooling inputs, given the level or distribution of funds available to schools background characteristics on student populations and districts, and for altering in substantive ways the natural variation (inequity) of the system.

PAGE 22 Studies relying on experiments implementation of specific comprehensive school Does Money reform models154 on student outcomes. Randomized Matter in Finally, there are those studies that rely on what trials are useful for studying specific reforms or Education? is considered the gold standard for research and models that may have cost implications, but, to the evaluation of educational programs—experimental best of my knowledge, randomized trials have not design studies. Experimental design studies been conducted to discern the importance of financial randomly assign one group of students to receive inputs to schooling directly, in part because doing so a specific set of programs and services and another would severely deprive some students of resources, group of students to a control group, or one that which would likely be objectionable to institutional does not receive the treatment of interest. Large- review boards and the general citizenry. Though, scale experimental design studies have been arguably, permitting the persistence of extreme conducted to determine the effects of class size natural variations is no less objectionable. reduction on student outcomes, participation in preschool programs on student outcomes, and the

Endnotes

1 For a comprehensive review of research on the 6 http://blogs.wsj.com/metropolis/2014/02/11/ effectiveness of early childhood education programs, cuomo-on-education-funding-lawsuit-it-aint- see W. S. Barnett, “Effectiveness of Early Educational about-the-money. Intervention,” Science 333, no. 6045 (2011): 975- 978. 7 http://online.wsj.com/article/SB10001424052702 303348504575184120546772244.html. 2 http://nces.ed.gov/programs/digest/d06/tables/ dt06_061.asp. 8 http://thinkprogress.org/education/ 2015/02/13/3623158/brat-education-plato. 3 For a version of this argument, see www. huffingtonpost.com/bill-gates/bill-gates- 9 B. D. Baker, “Evaluating the Recession’s Impact on school-performance_b_829771.html. However, State School Finance Systems,” Education Policy most characterizations of the extent of national Analysis Archives 22, no. 91 (2014), http://dx.doi. average spending increase are grossly oversimplified. org/10.14507/epaa.v22n91.2014. Further, on average, student achievement on the National Assessment of Educational Progress has 10 M. Leachman and C. Mai, Most States Still Funding actually improved quite dramatically over time and Schools Less Than Before the Recession, Center achievement gaps have narrowed. For example, on Budget and Policy Priorities, October 16, 2014, adjusting spending growth only for traditional www.cbpp.org/research/most-states-still- inflation measures, which account for changes in funding-schools-less-than-before-the-recession. the prices of consumer goods, does not account for 11 V. Bandeira de Mello, G. Bohrnstedt, C. Blankenship (a) changes in competitive wages of nonteachers, and D. Sherman, Mapping State Proficiency which influence the ability of schools to recruit and Standards onto NAEP Scales: Results from the 2013 retain teachers and have far outpaced the consumer NAEP Reading and Mathematics Assessments (NCES price index; (b) changes in the range and level of 2015-046) (Washington, DC: National Center for outcomes desired of our students, which affect costs Education Statistics, 2015), http://nces.ed.gov/ significantly; and (c) changes in the demographics pubsearch. of the student population, which affect the cost of achieving outcome objectives. See www.epi.org/ 12 Baker and Welner explain how the U.S. Department publication/fact-challenged_policy and www.ets. of Education has recently established a website on org/Media/Research/pdf/PICBWGAP.pdf. improving educational productivity, with specific intent to inform state policy and local practice. But, 4 http://blogs.orlandosentinel.com/news_ as Baker and Welner note: “None of the materials politics/2011/10/scott-anthropology-and- listed or recommendations expressed within those journalism-dont-pay-and-neither-do-capes.html. materials are backed by substantive analyses of the 5 http://www.governor.ny.gov/sl2/ cost effectiveness or efficiency of public schools, of stateofthestate2011transcript. practices within public schools, of broader policies pertaining to public schools, or of resource allocation

PAGE 23 Does Money strategies. Instead, the sources listed on the site’s (1968): 24. Matter in ‘resources’ page are speculative think tank reports Education? and related documents that do not include or even 16 S. Konstantopolous and G. Borman, “Family Background cite the types of analyses that would need to be and School Effects on Student Achievement: A Multilevel conducted to arrive at their conclusions and policy Analysis of the Coleman Data,” Teachers College Record recommendations.” B. D. Baker and K. G. Welner, 113, no. 1 (2011): 97-132. Evidence and Rigor: A Call for the U.S. Department of Education to Embrace High Quality Research 17 G. D. Borman and M. Dowling, “Schools and (National Education Policy Center, 2011). Inequality: A Multilevel Analysis of Coleman’s Equality of Educational Opportunity Data,” Teachers College 13 The scope of this review is limited to domestic Record 112, no. 5 (2010): 1201-1246. studies. The emphasis of the review is on major peer-reviewed studies in each of the three categories 18 D. Card and A. Krueger, “Does School Quality listed. Further, the emphasis is on studies that Matter? Returns to Education and the Characteristics use data aggregated to no higher level than local of Schools in the United States,” Journal of Political public school districts. That is, no cross-state or Economy 100, no. 1 (1992): 1-40. In a paper from cross-national aggregate analyses are emphasized, a few years later, Card and Krueger present a more though some are listed to point out their existence. tentative position on whether schooling resources For the older production function literature, the are clearly linked to earnings and attainment, a more bulk of the discussion herein focuses on major specific question. They note: “Does the literature on meta-analyses published in the late 1980s to late school resources, earnings and educational attainment 1990s, which reviewed studies from prior years. I prove beyond a reasonable doubt that resources do not reinvestigate those prior studies but do refer matter? We do not believe that the evidence justifies to some throughout. This review contains only a so strong a conclusion. The available evidence is not selected summary of major works on topics such unambiguous or ubiquitous, and it suffers from all as class size and teacher characteristics that have the standard criticisms of drawing causal inferences financial implications. For studies of state school from observational data.” See D. Card and A. finance reforms to be included, the studies must Krueger, “School Resources and Student Outcomes: measure more than the mere presence or nominal An Overview of the Lliterature and New Evidence indication that reform happened. Further, studies are from North and South Carolina,” Journal of Economic addressed if they attempt to measure the relationship Perspectives 10, no. 4 (1996): 31-50. between changes in the level of financial resources 19 J. Betts, “Is There a Link between School Inputs and for students in particular settings (districts, schools) Earnings? Fresh Scrutiny of an Old Literature,” in and subsequent changes in the level of outcomes, or Gary Burtless, ed., Does Money Matter? The Effect changes in the distribution of schooling resources of School Resources on Student Achievement and and distribution of student outcomes. Only studies Adult Success (Washington, DC: Brookings Institution, using short-term academic outcomes (measured 1996), 141-191. achievement, aptitude and graduation rates) are included. Preference is given to peer-reviewed studies 20 A later article by Hanushek, reiterating and updating and studies attempting to validate statistically a his earlier findings, also shows up as widely cited in link between changes in the level or distribution of the Social Science Citation Index: E. A. Hanushek, funding and the level or distribution of outcomes. “Assessing the Effects of School Resources on Student Performance: An Update,” Educational Evaluation 14 J. S. Coleman, E. Q. Campbell, C. F. Hobson, J. and Policy Analysis 19, no. 2 (1997): 141-164. McPartland, A. M. Mood, et al., Equality of Educational Opportunity (Washington, DC: U.S. 21 E. A. Hanushek, “Economics of Schooling: Production Office of Education, 1966). and Efficiency in Public Schools,” Journal of Economic Literature 24, no. 3 (1986): 1141-1177. A few 15 For an early discussion of the Coleman findings years later, Hanushek paraphrased this conclusion and misinterpretations of those findings with in another widely cited article as “Variations in respect to policy implications, see G. C. Kain and school expenditures are not systematically related to H. W. Watts, “Problems in Making Policy Inferences variations in student performance.” E. A. Hanushek, from the Coleman Report,” American Sociological “The Impact of Differential Expenditures on School Review 35, no. 2 (1970): 228-242; and S. Bowles Performance,” Educational Researcher 18, no. 4 and H. M. Levin, “The Determinants of Scholastic (1989): 45-62. Hanushek describes the collection Achievement—an Appraisal of Some Recent of studies relating spending and outcomes as Evidence,” Journal of Human Resources 3, no. 1 follows: “The studies are almost evenly divided

PAGE 24 between studies of individual student performance Developments in School Finance, 1997, ed. W. J. Does Money and aggregate performance in schools or districts. Fowler Jr. (Washington, DC: U.S. Department of Matter in Ninety-six of the 147 studies measure output by Education, National Center for Education Statistics, Education? score on some standardized test. Approximately 40 1998), http://nces.ed.gov/pubs98/98212. percent are based upon variations in performance pdf#page=83. within single districts while the remainder look across districts. Three-fifths look at secondary performance 26 B. D. Baker, “Can Flexible Non-Linear Modeling Tell (grades 7-12) with the rest concentrating on Us Anything New about Educational Productivity?,” elementary student performance” (fn #25). Economics of Education Review 20, no. 1 (2001): 81-92; D. N. Figlio, “Functional Form and the 22 Greenwald and colleagues explain: “Studies in the Estimated Effects of School Resources,” Economics of universe Hanushek (1989) constructed were assessed Education Review 18, no. 2 (1999): 242-252; and J. for quality. Of the 38 studies, 9 were discarded Dewey, T. Husted and L. Kenny, “The Ineffectiveness due to weaknesses identified in the decision rules of School Inputs: A Product of Misspecification,” for inclusion described below. While the remaining Economics of Education Review 19, no. 1 (2000): 29 studies were retained, many equations and 27-45. coefficients failed to satisfy the decision rules we employed. Thus, while more than three quarters of 27 Specifically, Dewey and colleagues explain that the studies were retained, the number of coefficients many previous studies attempting to distill school from Hanushek’s universe was reduced by two thirds” resource effects on student outcomes concurrently (p. 363). Greenwald and colleagues further explain correct for the economic background of students. that: “Hanushek’s synthesis method, vote counting, However, the economic background measures, such consists of categorizing, by significance and direction, as family income, are also strong determinants of the relationships between school resource inputs the demand for schooling resources. Thus, including and student outcomes (including but not limited to the two simultaneously in regression models violates achievement). Unfortunately, vote-counting is known both conceptual appropriateness (resource levels are to be a rather insensitive procedure for summarizing endogenous to family characteristics) and statistical results. It is now rarely used in areas of empirical properties associated with those conceptual problems research where sophisticated synthesis of research is (that the error term is correlated with the school expected” (p. 362). input measures, requiring a different statistical approach). Dewey and colleagues review the previous studies summarized by Hanushek, identifying that Hanushek (1997) provides his rebuttal to some several suffer from this problem and that those that of these arguments, and Hanushek returns to his do tend to understate the influence of resources. “uncertainty” position: “The close to 400 studies of Then Dewey and colleagues estimate alternative student achievement demonstrate that there is not production functions. a strong or consistent relationship between student performance and school resources, at least after We conducted our own empirical analysis using the variations in family inputs are taken into account” Project TALENT student-level data set from 1960 (p. 141). E. A. Hanushek, “Assessing the Effects and pooled state data for 1987-1992. In regressions of School Resources on Student Performance: An from both data sets that were not plagued by Update,” Educational Evaluation and Policy Analysis misspecification, there is evidence that each school 19, no. 2 (1997): 141-164. See also E. A. Hanushek, input had an impact on achievement (p. 42). Figlio’s “Money Might Matter Somewhere: A Response study of alternative specifications of the “shape” of to Hedges, Laine and Greenwald,” Educational the relationship between money and outcomes raises Researcher 23 (May 1994): 5-8. similar issues about previous literature, including studies summarized by Hanushek, as does Corrine 23 R. Greenwald, L. Hedges and R. Laine, “The Effect of Taylor’s analysis that applies adjustments for the School Resources on Student Achievement,” Review costs of hiring teachers. Indeed, many of the same of Educational Research 66, no. 3 (1996): 361-396. studies considered rigorous enough for inclusion in Greenwald and colleagues analyses also suffer from 24 H. Wenglinsky, “How Money Matters: The Effect of the problems addressed by Husted and Kenny, and School District Spending on Academic Achievement,” by Taylor (geographic cost adjustment) and Figlio. Sociology of Education 70, no. 3 (1997): 221-237. But, note that in each case, Dewey and colleagues, Taylor and Figlio find that when applying functional 25 C. Taylor. “Does Money Matter? An Empirical Study form and labor cost corrections, they tend to find Introducing Resource Costs and Student Needs stronger effects of schooling resources—specifically into Educational Production Function Analysis,” in

PAGE 25 Does Money money. So, one might then argue that Greenwald (unintentionally). The endogeneity problems are less Matter in and colleagues decisively positive findings are, in clear in this study, because in place of controlling Education? fact, understated. In conducting this review, I went for direct demand determinants (family income, back to a handful of the original studies summarized education), the authors control for individual IQ. by Hanushek (1986) and listed in the sources note However, IQ is arguably simultaneously determined to Table 8 of that article. Several were not easily with education spending, both IQ and school accessible, having been non-peer-reviewed reports spending being a function of parental economic and doctoral theses. But among those available, status and education level. Sensitive to this point, the consistent with the findings of Husted and Kenny, authors explore direct and indirect effects of IQ, years none attempted to account for the endogeneity of education and expenditures. of expenditures, often either evaluating simple R. Raymond, “Determinants of the Quality of correlations between spending and outcome Primary and Secondary Public Education in West measures (thus suffering significant omitted variables Virginia,” Journal of Human Resources 3, no. 4 bias) or including a spending measure alongside (1968): 450-470. Raymond studied 5,000 students determinants of spending. Arguably, teacher in West Virginia. Raymond did not explore per-pupil characteristics, including teacher salaries, are also expenditures but did explore several teacher salary endogenous to local demand factors. Original measures, but does not correct for regional variation Hanushek studies reviewed include: in the value of those salaries across West Virginia. A. Boardman, O. Davis and P. Sanday, “A Raymond finds salaries to be associated with output Simultaneous Equations Model of the Educational measures of quality. Process,” Journal of Public Economics 7, no. 1 T. I. Ribich and J. L. Murphy, “The Economic (1977): 23-49. This study does not explore Returns to Increased Educational Spending,” Journal expenditures directly but does include measures of of Human Resources 10, no. 1 (1975): 56-77. Ribich schooling facilities and teacher characteristics, but and Murphy used data from the national Project not salary. Thus, regional cost variation is less (or Talent survey. Ribich and Murphy found: “School not) for the value of teacher salaries or education expenditures are found to influence how many spending is less at issue. The authors of this years of schooling an individual eventually receives, study find that “many educational outputs jointly and the chief effect of spending differences on determine one another. Also, the results suggest that lifetime income is found to work through this school school and teacher variables have important effects continuation link” (p. 56). Ribich and Murphy partly on educational outcomes” (p. 23). (though far from completely) correct for regional G. E. Johnson and F. P. Stafford, “Social Returns differences in the value of expenditures by including to Quantity and Quality of Schooling,” Journal of region variables. But, regression estimates likely suffer Human Resources 8, no. 2 (1973): 139-155. In endogeneity addressed by Dewey, Husted and Kenny this study, the authors find “high but diminishing (including both family socioeconomic measures and marginal returns to investment in expenditures expenditures alongside one another). Interestingly, per pupil per year” (p. 139). This is among the the authors instead attribute the insensitivity studies that arguably understates the sensitivity of their outcome measures to spending (when of expenditures to outcomes by inclusion of the directly estimated including all regions) to regional spending measure (natural log of expenditures) in differences, specifically racial differences within the model with determinants of expenditure (family southern states. socioeconomic status). In addition, the model uses F. Welch, “Measurement of the Quality of a national sample but fails to control for regional Schooling,” American Economic Review 56 variation in the value of expenditures. (1966): 379-392. This study explored the return to C. R. Link and E. C. Ratledge, “Social Returns elementary and secondary schooling of the male to Quantity and Quality of Education: A Further rural farm population in 1959, focusing on those Statement,” Journal of Human Resources 10, no. 1 who had not attended college in an effort to isolate (1975): 78-89. Link and Ratledge find: “Large but differences in elementary and secondary schooling diminishing returns to incremental expenditures quality. This study is problematic on a number of are observed” (p. 78). Link and Ratledge also use levels when viewed in hindsight. First, the ultimate national survey data (National Longitudinal Study analysis of factors associated with the quality of of the Labor Force). For the expenditure measure, schooling is aggregated to the state level (and noted like the above study, they use a measure of the by the author as a significant limitation). Second, 1968 district level per-pupil expenditures (natural expenditure measures are included in models with logarithm) and also do not correct for regional (a) potential determinants of expenditures (racial variation, though some of the urbanicity variables composition, labor composition, enrollment per included may capture a portion of this variation secondary school); and (b) schooling resources

PAGE 26 dependent on expenditures (salaries, staff per 100 Longitudinal Analysis with Student Fixed Effects,” Does Money pupils) (see regression output in Table 4, p. 390). Economics of Education Review 26 (2007): 673-682; Matter in Further, expenditures are not adjusted for regional D. Goldhaber and D. Brewer, “Why Don’t Schools Education? differences in value, nor are salaries. and Teachers Seem to Matter? Assessing the Impact of Unobservables on Educational Productivity,” 28 In tangentially related work, Hanushek, Rivkin and Journal of Human Resources 332, no. 3 (1997): 505- Taylor (1996) explore the influence of aggregation 523; R. G. Ehrenberg and D. J. Brewer, “Do School bias and omitted variables on estimates of the and Teacher Characteristics Matter? Evidence from relationship between teacher characteristics and High School and Beyond,” Economics of Education student outcomes, using data from the High School Review 13, no. 1 (1994): 1-17; R. G. Ehrenberg and and Beyond survey. They find that at higher levels of D. J. Brewer, “Did Teachers’ Verbal Ability and Race aggregation, studies tend to overstate the strength Matter in the 1960s?,” Economics of Education of the relationship between resources and student Review 14, no. 1 (1995): 1-21; C. Jepsen, “Teacher outcomes, but raise the most significant concerns Characteristics and Student Achievement: Evidence about studies using data aggregated to the state from Teacher Surveys,” Journal of Urban Economics level with crude aggregate state level measures of 57, no. 2 (2005): 302-319; B. A. Jacob and L. student and population characteristics, far beyond Lefgren, “The Impact of Teacher Training on Student the aggregation of most recent studies. Achievement: Quasi-Experimental Evidence from E. A. Hanushek, S. Rivkin and L. L. Taylor, School Reform,” Journal of Human Resources 39, “Aggregation Bias and the Estimated Effects of School no. 1 (2004): 50-79; S. G. Rivkin, E. A. Hanushek Resources,” Review of Economics and Statistics 78, and J. F. Kain, “Teachers, Schools, and Academic no. 4 (1996): 611-627. Along these lines, there does Achievement,” Econometrica 73, no. 2 (2005): exist a separate body of literature that endeavors to 471; and A. J. Wayne and P. Youngs, “Teacher prove that education spending is not associated with Characteristics and Student Achievement Gains,” student outcomes by making national aggregate Review of Educational Research 73, no. 1 (2003): comparisons of spending and outcomes. That is, 89-122. For a recent review of studies on the returns by showing that on average, countries that spend to teacher experience, see J. K. Rice, The Impact more per pupil don’t perform better on international of Teacher Experience: Examining the Evidence and assessments. See, for example Policy Implications (National Center for Analysis of H. J. Walberg, “Spending More While Learning Longitudinal Data in Educational Research, 2010). Less,” Fordham Report 2, no. 6 (1998). These studies suffer sufficiently from aggregation issues to be of 31 Some go so far as to argue that half or more little importance to the discussion herein. While of teacher pay is allocated to “non-productive” aggregation might lead to overstating the money- teacher attributes, and so it follows that that outcome relationship in some studies, these studies entire amount of funding could be reallocated also suffer from numerous substantial measurement toward making schools more productive. See, for problems regarding both input and outcome example, a recent presentation to the New York measures. For example, education spending data are State Board of Regents from Sept. 13, 2011 (page simply not directly comparable across nations, partly 32), slides by Stephen Frank of Education Resource because they include vastly different programs and Strategies: www.p12.nysed.gov/mgtserv/docs/ services (athletics, arts, special education) as well as SchoolFinanceForHighAchievement.pdf. other specific expenses, such as health insurance costs for U.S. school employees, that may be covered 32 H. Lankford, S. Loeb and J. Wyckoff, “Teacher via other government programs in other nations. Sorting and the Plight of Urban Schools,” Educational Evaluation and Policy Analysis 24 (2002): 29 Hanushek (1986) explains: “Thus the basic 37-62. determinants of instructional expenditures in a district are teacher experience, teacher education and 33 S. A. Allegretto, S. P. Corcoran and L. R. Mishel, class size, and most studies, regardless of what other The Teaching Penalty: Teacher Pay Losing Ground descriptors of schools might be included, will analyze (Washington, DC: Economic Policy Institute, 2008). the effect of these factors on outcomes” (p. 1160). 34 R. J. Murnane and R. Olsen, “The Effects of Salaries 30 E. A. Hanushek, “Teacher Characteristics and and Opportunity Costs on Length of Stay in Teaching: Gains in Student Achievement: Estimation Using Evidence from Michigan,” Review of Economics MicroData,” Econometrica 61, no. 2 (1971): 280- and Statistics 71, no. 2 (1989): 347-352. 288; C. T. Clotfelter, H. F. Ladd and J. L. Vigdor, “Teacher Credentials and Student Achievement: 35 D. N. Figlio, “Can Public Schools Buy Better-Qualified

PAGE 27 Does Money Teachers?,” Industrial and Labor Relations Review Experiment with Schoolwide Performance Bonuses, Matter in 55 (2002): 686-699; D. N. Figlio “Teacher Salaries Final Evaluation Report (RAND Corporation and Education? and Teacher Quality,” Economics Letters 55 (1997): Vanderbilt University, 2011). 267-271; and R. Ferguson, “Paying for Public Education: New Evidence on How and Why Money 42 K. Yuan, V. N. Le, D. F. McCaffrey, J. A. Marsh, L. S. Matters,” Harvard Journal on Legislation 28, no. 2 Hamilton, B. M. Stecher and M. G. Springer, “Incentive (1991): 465-498. Pay Programs Do Not Affect Teacher Motivation or Reported Practices: Results from Three Randomized 36 S. Loeb and M. Page, “Examining the Link Studies,” Educational Evaluation and Policy Analysis Between Teacher Wages and Student Outcomes: 35, no. 1 (2013); M. G. Springer, D. Ballou, L. Hamilton, The Importance of Alternative Labor Market V. N. Le, J. R. Lockwood, D. F. McCaffrey and B. M. Opportunities and Non-Pecuniary Variation,” Review Stecher, “Teacher Pay for Performance: Experimental of Economics and Statistics 82, no. 3 (2000): 393- Evidence from the Project on Incentives in Teaching 408. (POINT),” Society for Research on Educational Effectiveness (2011); and M. G. Springer, J. F. Pane, V. 37 D. N. Figlio and K. Rueben, “Tax Limits and the N. Le, D. F., McCaffrey, S. F. Burns, L. S. Hamilton and Qualifications of New Teachers,” Journal of Public B. Stecher, “Team Pay for Performance Experimental Economics (April 2001): 49-71. See also T. A. Evidence from the Round Rock Pilot Project on Team Downes and D. N. Figlio, “Do Tax and Expenditure Incentives,” Educational Evaluation and Policy Analysis Limits Provide a Free Lunch? Evidence on the Link 34, no. 4 (2012): 367-390. Between Limits and Public Sector Service Quality,” National Tax Journal 52, no. 1 (1999): 113-128. 43 T. S. Dee and J. Wyckoff, “Incentives, Selection, and Teacher Performance: Evidence from IMPACT,” 38 J. Ondrich, E. Pas and J. Yinger, “The Determinants Journal of Policy Analysis and Management 34, no. 2 of Teacher Attrition in Upstate New York,” Public (2015): 267-297. Finance Review 36, no. 1(2008): 112-144. 44 R. Balch and M. G. Springer, “Performance Pay, Test 39 E. A. Hanushek, J. F. Kain and S. G. Rivkin, “Why Scores, and Student Learning Objectives,” Economics Public Schools Lose Teachers,” Journal of Human of Education Review 44 (2015): 114-125. Resources 39, no. 2 (2004): 350. 45 A. J. Sojourner, E. Mykerezi and K. L. West, “Teacher Pay Reform and Productivity Panel Data Evidence from Adoptions of Q-Comp in Minnesota,” Journal 40 C. Clotfelter, H. F. Ladd and J. Vigdor, “Teacher of Human Resources 49, no. 4 (2014): 945-981. Mobility, School Segregation and Pay-Based Policies to Level the Playing Field,” Education Finance and 46 R. G. Fryer Jr., S. D. Levitt, J. List and S. Sadoff, Policy 6, no. 3 (2011): 399-438; and C. T. Clotfelter, Enhancing the Efficacy of Teacher Incentives through E. Glennie, H. F. Ladd and J. L. Vigdor, “Would Loss Aversion: A Field Experiment (no. w18237) Higher Salaries Keep Teachers in High-Poverty (National Bureau of Economic Research, 2012). Schools? Evidence from a Policy Intervention in North Carolina,” Journal of Public Economics 92 (2008): 47 M. D. Hendricks, Public Schools Are Hemorrhaging 1352-1370. Talented Teachers: Can Higher Salaries Function as a Tourniquet?, March 24, 2015, http://papers.ssrn. 41 For major studies specifically on the topic of “merit com/sol3/papers.cfm?abstract_id=2564703. pay,” each of which generally finds no positive effects of merit pay on student outcomes, see S. 48 M. D. Hendricks, “Towards an Optimal Teacher Salary Glazerman and A. Seifullah, An Evaluation of the Schedule: Designing Base Salary to Attract and Retain Teacher Advancement Program in Chicago: Year Two Effective Teachers,” Economics of Education Review Impact Report (Mathematica Policy Research Institute, (2015), doi: 10.1016/j.econedurev.2015.05.008. 2010); M. G. Springer, D. Ballou, L. Hamilton, V. Le, J. R. Lockwood, D. McCaffrey, M. Pepper and B. 49 For a review of much the same literature, see www. Stecher, Teacher Pay for Performance: Experimental shankerinstitute.org/blog/recent-evidence- Evidence from the Project on Incentives in Teaching teacher-experience-and-productivity. (Nashville, TN: National Center on Performance Incentives at Vanderbilt University, 2010); and J. A. 50 M. Wiswall, “The Dynamics of Teacher Quality,” Marsh, M. G. Springer, D. F. McCaffrey, K. Yuan, S. Journal of Public Economics 100 (2013): 61-78. Epstein, J. Koppich, N. Kalra, C. DiMartino and A. 51 J. P. Papay and M. A. Kraft, “Productivity Returns Peng, A Big Apple for Educators: New York City’s

PAGE 28 to Experience in the Teacher Labor Market: no. 1 (November 2009): 125-154. Does Money Methodological Challenges and New Evidence on Matter in Long-Term Career Improvement,” Journal of Public 57 A. Krueger, “Experimental Estimates of Education Education? Economics (2015). Production Functions,” Quarterly Journal of Economics 114, no. 2 (1999): 497-532. 52 H. F. Ladd and L. C. Sorensen, Returns to Teacher Experience: Student Achievement and Motivation 58 S. Konstantopoulos and V. Chun, “What Are in Middle School, Working Paper 112 (National the Long-Term Effects of Small Classes on the Center for Analysis of Longitudinal Data in Education Achievement Gap? Evidence from the Lasting Research, 2014). Benefits Study,” American Journal of Education 116, no. 1 (November 2009): 125-154. 53 J. Rothstein, “Teacher Quality Policy When Supply Matters,” Document de treball de IEB 35 (2012): 59 S. Dynarski, J. Hyman and D. W. Schanzenbach, 1-65; and J. Rothstein, Teacher Quality Policy When “Experimental Evidence on the Effect of Childhood Supply Matters (no. w18419) (National Bureau of Investments on Postsecondary Attainment and Economic Research, 2012). Degree Completion,” Journal of Policy Analysis and Management 32, no. 4 (2013): 692-717. 54 www.americanprogress.org/issues/2011/04/pdf/ class_size.pdf. 60 Another relevant study showing positive effects of pupil-to-teacher ratio reduction (different from class 55 See, for example, D. J. Brewer, C. K. Kropp, B. size) is the Wisconsin SAGE study. See A. Molnar, P. P. Gill and R. Reichardt, “Estimating the Cost of Smith, J. Zahorik, A. Palmer, A. Halbach and K. Ehrle, National Class Size Reductions Under Different Policy “Evaluating the SAGE Program: A Pilot Program in Alternatives,” Educational Evaluation and Policy Targeted Pupil-Teacher Reduction in Wisconsin,” Analysis 21, no. 2 (1999): 171-192. While this article Educational Evaluation and Policy Analysis 21, no. provides insights into the cumulative costs of adding 2 (1999): 165-177. Unlike STAR, which was a large numbers of teachers, it makes no comparisons true randomized experiment in Tennessee, SAGE to other strategies that might be employed for the in Wisconsin was designed as “a 5-year K-3 pilot same dollar. The article acknowledges the research project that began in the 1996-97 school year. on positive effects of class size and then estimates The program required that participating schools large-scale implementation costs, seemingly implying implement 4 interventions including reducing the either that achieving these positive effects is simply pupil-teacher ratio within classrooms to 15 students too expensive or that there might be more cost- per teacher” (p. 165). Molnar and colleagues found: effective uses of the same dollar. “Results of the 1996-97 and 1997-98 first grade data reveal findings consistent with the Tennessee STAR 56 See http://www2.ed.gov/rschstat/research/pubs/ class size experiment” (p. 165). For an example of a rigorousevid/rigorousevid.pdf; J. D. Finn and C. study based on natural variation, finding no positive M. Achilles, “Tennessee’s Class Size Study: Findings, effects of smaller class size, see C. M. Hoxby, “The Implications, Misconceptions,” Educational Evaluation Effects of Class Size on Student Achievement: New and Policy Analysis 21, no. 2 (Summer 2009): Evidence from Population Variation,” Quarterly 97-109; J. Finn et al., “The Enduring Effects of Small Journal of Economics 115, no. 4 (2000): 1239-1285. Classes,” Teachers College Record 103, no. 2 (April Hoxby uses grade-level, not student-level, data on 2001): 145-183, www.tcrecord.org/pdf/10725. 649 elementary schools in Connecticut, concluding pdf; A. Krueger, “Would Smaller Class Sizes Help “class size does not have a statistically significant Close the Black-White Achievement Gap,” Working effect on student achievement” (p. 1239). Paper 451 (Princeton, NJ: Industrial Relations Section, Department of Economics, Princeton University, 61 M. M. Chingos, “The False Promise of Class-Size 2001), www.irs.princeton.edu/pubs/working_ Reduction” (Center for American Progress, 2011), papers.html; H. M. Levin, “The Public Returns to https://cdn.americanprogress.org/wp-content/ Public Educational Investments in African American uploads/issues/2011/04/pdf/class_size.pdf. Males” (Dijon Conference, University of Bourgogne, France, May 2006), http://www.u-bourgogne.fr/ 62 M. M. Chingos, “Class Size and Student Outcomes: colloque-iredu/posterscom/communications/ Research and Policy Implications,” Journal of Policy LEVIN.pdf; and S. Konstantopoulos and V. Chun, Analysis and Management 32, no. 2 (2013): 411-438. “What Are the Long-Term Effects of Small Classes on the Achievement Gap? Evidence from the Lasting 63 Referring to estimates provided by A. B. Krueger, Benefits Study,” American Journal of Education 116, “Experimental Estimates of Education Production Functions,” Quarterly Journal of Economics 115

PAGE 29 Does Money (1999): 497-532. See also H. Cho, P. Glewwe and M. Saez, D. W. Schazenbach and D. Yagan, “How Does Matter in Whitler, “Do Reductions in Class Size Raise Students’ Your Kindergarten Classroom Affect Your Earnings? Education? Test Scores? Evidence from Population Variation in Evidence from Project STAR,” NBER Working Paper Minnesota’s Elementary Schools,” Economics of 16381 (Cambridge, MA: National Bureau of Economic Education Review 31 (2012): 77-95. Research, 2010), www.nber.org/papers/w16381; P. Blatchford, P. Bassett and P. Brown, “Teachers’ 64 Head Start costs $8,000 per child. Given the 6 and Pupils’ Behavior in Large and Small Classes: A percentage-point effect noted above, the amount Systematic Observation Study of Pupils Aged 10 and spent by Head Start to induce a single child into 11 Years,” Journal of Educational Psychology 97, college is therefore $133,333 ($8,000/0.06). For no. 3 (August 2005): 454-467, doi: 10.1037/0022- Abecedarian, the figure is $410,000 ($90,000/0.22). 0663.97.3.454; P. Babcock and J. Betts, “Reduced The cost of reduced class size is $12,000 per student, Class Size Distinctions: Effort, Ability and the larger than that of Head Start but considerably Education Production Function,” NBER Working smaller than that of Abecedarian. The amount Paper 14777 (Cambridge, MA: National Bureau of spent in STAR to induce a single child into college Economic Research, 2009), www.nber.org/papers/ is $400,000 ($12,000/0.03). If the program could w14777; and S. T. Lubienski, C. Lubienski and C. be focused on students in the poorest third of Crawford-Crane, “Achievement Differences and schools (the subpopulation that most closely matches School Type: The Role of School Climate, Teacher that of the preschool interventions), then the Certification, and Instruction,” American Journal of cost drops to $171,000 per student induced into Education 115 (2008): 97-138. college. Upward Bound costs $5,620 per student. If the program could be targeted to students with 67 C. Jepsen and S. Rivkin, “What Is the Tradeoff low educational aspirations, the implied cost of between Smaller Classes and Teacher Quality?,” inducing a single student into college is $93,667 NBER Working Paper 9205 (Cambridge, MA: ($5,620/0.06). Dynarski (2003) examines the effect National Bureau of Economic Research, 2002), www. of the elimination of the Social Security Student nber.org/papers/w9205. “The results show that, Benefit Program, which paid college scholarships to all else being equal, smaller classes raise third-grade the dependents of deceased, disabled and retired mathematics and reading achievement, particularly Social Security beneficiaries. Eligible students were for lower-income students. However, the expansion disproportionately black and low-income. The of the teaching force required to staff the additional estimates from that paper indicate that about two- classrooms appears to have led to a deterioration thirds of the treated students who attended college in average teacher quality in schools serving a were inframarginal, while the other third was predominantly black student body. This deterioration induced into the college by the $7,000 scholarship. partially or, in some cases, fully offset the benefits These estimates imply that three students are paid a of smaller classes, demonstrating the importance scholarship in order to induce one into college. The of considering all implications of any policy change” cost per student induced into college is therefore (p. 1). For further discussion of the complexities of $21,000. Finally, the cost per treated subject in the evaluating class size reduction in a dynamic policy FAFSA experiment (Bettinger et al., 2012) was $88, context, see D. Sims, “A Strategic Response to Class for an implied cost per student induced into college Size Reduction: Combination Classes and Student of $1,100 ($88/0.08). S. Dynarski, J. M. Hyman and Achievement in California,” Journal of Policy Analysis D. W. Schanzenbach, Experimental Evidence on the and Management 27, no. 3 (2008): 457-478; D. Effect of Childhood Investments on Postsecondary Sims, “Crowding Peter to Educate Paul: Lessons Attainment and Degree Completion (no. w17533) from a Class Size Reduction Externality,” Economics (National Bureau of Economic Research, 2011). of Education Review 28 (2009): 465-473; and M. M. Chingos, “The Impact of a Universal Class-Size 65 Including recent work linking participation in smaller Reduction Policy: Evidence from Florida’s Statewide class sizes with postsecondary degree attainment: Mandate,” Program on Education Policy and S. Dynarski, J. M. Hyman and D. W. Schazenbach, Governance Working Paper 10-03 (2010). “Experimental Evidence on the Effect of Childhood Investments on Postsecondary Attainment and 68 R. G. Ehrenberg, D. Brewer, A. Gamoran and J. D. Degree Completion,” NBER Working Paper 17533 Willms, “Class Size and Student Achievement,” (Cambridge, MA: National Bureau of Economic Psychological Science in the Public Interest 2, no. 1 Research, 2011), www.nber.org/papers/w17533. (2001): 1-30.

66 For other relatively recent studies on class size 69 See B. D. Baker, D. Farrie, D. Sciarra and R. reduction, see R. Chetty, J. N. Friedman, N. Hilger, E. Coley, “Is School Funding Fair,” 2010, www.

PAGE 30 schoolfundingfairness.org. that Neymotin evaluated changes in school funding Does Money from 1997 to 2006, but the first additional funding Matter in 70 Two reports from Cato are illustrative: P. Ciotti, infused following the January 2005 Supreme Court Education? Money and School Performance: Lessons from the decision occurred in the 2005-06 school year, the Kansas City Desegregations Experience, Cato Policy end point of Neymotin’s outcome data. F. Neymotin, Analysis 298 (1998); and D. Coate and J. VanderHoff, “The Relationship between School Funding and “Public School Spending and Student Achievement: Student Achievement in Kansas Public Schools,” The Case of New Jersey,” Cato Journal 19, no. 1 Journal of Education Finance 36, no. 1 (2010): (1999): 85-99. Edspresso, New Jersey Learns Kansas 88-108. City’s Lessons the Hard Way (2006), retrieved Oct. 23, 2009, from www.edspresso.com/index. 73 In an earlier, edited volume, Hanushek goes so far php/2006/10/new-jersey-learns-kansas-citys- as to title the book “How School Finance Lawsuits lessons-the-hard-way-2. Exploit Judges’ Good Intentions and Harm Our Children” (emphasis added). The premise that 71 E. A. Hanushek and A. Lindseth, Schoolhouses, additional funding for schools, often leveraged Courthouses and Statehouses (Princeton, NJ: toward class size reduction, additional course Princeton University Press, 2009). See also http:// offerings or increased teacher salaries, causes edpro.stanford.edu/Hanushek/admin/pages/ harm to children is, on its face, absurd. And the files/uploads/06_EduO_Hanushek_g.pdf. book, which implies as much in its title, never once validates that such reforms ever do cause harm. 72 Kevin Welner and I explain that Hanushek and Rather, the title is little more than a manipulative Lindseth failed to even measure whether substantive attempt to convince the noncritical spectator who changes had occurred to the level or distribution never gets past the book’s cover to fear that school of school funding as well as when and for how finance reforms might somehow harm children. long. We point out that in New Jersey, for example, That is, adding an element of fear to the cloud of infusion of funding occurred from 1998 to 2003 doubt. See, for example, E. A. Hanushek, Courting (or 2005), thus Hanushek and Lindseth’s window Failure: How School Finance Lawsuits Exploit includes six years on the front end where little Judges’ Good Intentions and Harm Our Children change occurred. Kentucky reforms had largely (Hoover Institution Press, 2006). A review of the faded by the mid- to late 1990s, yet Hanushek book is available at www.tcrecord.org/Content. and Lindseth measure postreform effects in 2007. asp?ContentId=13382. This book also includes Further, in New Jersey, funding was infused into two examples of a type of analysis that occurred approximately 30 specific districts, but Hanushek with some frequency in the mid-2000s that also and Lindseth explore overall changes to outcomes had the intent of showing that school funding among low-income children and minorities using doesn’t matter. These studies would cherry pick NAEP data, where some of these children attend anecdotal information from either or both (a) poorly the districts receiving additional support but many funded schools that have high outcomes; and (b) did not. In short, the slipshod comparisons made well-funded schools that have low outcomes. The by Hanushek and Lindseth provide no reasonable implication would be that if such schools exist, basis for asserting either the success or failure of money must not matter. See W. M. Evers and P. state school finance reforms. We also discuss other Clopton, “High-Spending, Low-Performing School studies that involve similar flaws of reasoning. For Districts,” in Courting Failure: How School Finance example, Greene and Trivitt present a study in which Lawsuits Exploit Judges’ Good Intentions and Harm they claim to show that court-ordered school finance Our Children, ed. E. A. Hanushek (Palo Alto, CA: reforms led to no substantive improvements in Hoover Press, 2006), 103-194; and H. Walberg, student outcomes. However, the authors test only “High-Poverty, High-Performance Schools, Districts whether the presence of a court order is associated and States,” in Courting Failure: How School Finance with changes in outcomes, and never once measure Lawsuits Exploit Judges’ Good Intentions and Harm whether substantive school finance reforms followed Our Children, ed. E. A. Hanushek (Palo Alto, CA: the court order. See J. P. Greene and J. R. Trivitt, Hoover Press, 2006), 79-102. “Can Judges Improve Academic Achievement?,” Peabody Journal of Education 83, no. 2 (2008): 224- 74 For additional discussion of the strengths and 237. In an equally problematic analysis, Florence weaknesses of this particular study, see Baker and Neymotin set out to show that massive court-ordered Welner (2011). infusions of funding in Kansas following Montoy v. Kansas led to no substantive improvements in 75 D. Card and A. A. Payne, “School Finance Reform, student outcomes. However, Welner and I explain the Distribution of School Spending, and the

PAGE 31 Does Money Distribution of Student Test Scores,” Journal of Public volume, by volume editor John Yinger: J. Yinger, Matter in Economics 83, no. 1 (2002): 49-82. See Baker and “State Aid and the Pursuit of Educational Equity: An Education? Welner (2011) for a more thorough discussion of Overview,” in Helping Children Left Behind: State Aid the Card and Payne analysis and its strengths and and the Pursuit of Educational Equity, ed. J. Yinger weaknesses. (MIT Press, 2004), 39.

76 The authors explain: “Our sample consists of PSID 81 J. Roy, “Impact of School Finance Reform on sample members born between 1955 and 1985 Resource Equalization and Academic Performance: who have been followed from 1968 into adulthood Evidence from Michigan,” Education Finance through 2011. This corresponds to cohorts that both and Policy 6, no. 2 (2011): 137-167. Roy (2011) straddle the first set of court-mandated SFRs (the first published an analysis of the effects of Michigan’s of which was in 1972) and are old enough to have school finance reforms of the 1990s, which led to completed formal schooling by 2011. Two-thirds a significant leveling up for previously low-spending of those in these cohorts in the PSID grew up in a districts. Roy, whose analyses measure both whether school district that was subject to a court-mandated the policy resulted in changes in funding and school finance reform between 1972 and 2000.” C. who was affected, found that “Proposal A was K. Jackson, R. C. Johnson and C. Persico, The Effects quite successful in reducing interdistrict spending of School Spending on Educational and Economic disparities. There was also a significant positive effect Outcomes: Evidence from School Finance Reforms on student performance in the lowest-spending (no. w20847) (National Bureau of Economic Research, districts as measured in state tests” (p. 137). 2015). 82 L. Papke, “The Effects of Spending on Test Pass 77 B. Baker and K. Welner, “School Finance and Rates: Evidence from Michigan,” Journal of Public Courts: Does Reform Matter, and How Can We Tell,” Economics 89, no. 5-6 (2005): 821-839; and J. Teachers College Record 113, no. 11 (2011): 2374- Hyman, Does Money Matter in the Long Run? Effects 2414. of School Spending on Educational Attainment (2013), www-personal.umich.edu/~jmhyman/ 78 http://educationnext.org/boosting-education- Hyman_JMP.pdf. attainment-adult-earnings-school-spending. 83 J. Roy, “Impact of School Finance Reform on 79 In fact, a table often cited by Hanushek to show Resource Equalization and Academic Performance: that investment in low-income schools has increased Evidence from Michigan,” Education Research over time shows that by 2006-07, lower-income Section Working Paper No. 8 (Princeton University, schools, nationally, were spending, on average, 2003), retrieved Oct. 23, 2009, from http://papers. slightly more than $100 per pupil more than higher ssrn.com/sol3/papers.cfm?abstract_id=630121 income schools. See Table A-36-1: S. Aud, W. Hussar, (forthcoming in Education Finance and Policy). M. Planty, T. Snyder, K. Bianco, M. Fox, L. Frohlich, J. Kemp and L. Drake, The Condition of Education 84 L. Papke, “The Effects of Spending on Test Pass 2010 (NCES 2010-028) (Washington, DC: National Rates: Evidence from Michigan,” Journal of Public Center for Education Statistics, Institute of Education Economics 89, no. 5-6 (2005): 821-839. In a Sciences, U.S. Department of Education, 2010). separate study, Leuven and colleagues attempted to This national average margin of about 1.2 percent isolate specific effects of increases to at-risk funding is hardly sufficient for eliminating all achievement on at-risk pupil outcomes but did not find any gaps, as explained by Duncombe and Yinger: W. positive effects. E. Leuven, M. Lindahl, H. Oosterbeek Duncombe and J. Yinger, “Why Is It So Hard to Help and D. Webbink, “The Effect of Extra Funding for Central City Schools?,” Journal of Policy Analysis Disadvantaged Pupils on Achievement,” Review of and Management 16, no. 1 (1997): 85-113; and W. Economics and Statistics 89, no. 4 (2007): 721-736. Duncombe and J. Yinger, “How Much More Does a Disadvantaged Student Cost?,” Economics of 85 J. Hyman, Does Money Matter in the Long Education Review 24, no. 5 (2005): 513-532. Run? Effects of School Spending on Educational Attainment (2013), www-personal.umich. 80 D. N. Figlio, “Funding and Accountability: Some edu/~jmhyman/Hyman_JMP.pdf. Conceptual and Technical Issues in State Aid Reform,” in Helping Children Left Behind: State Aid and the 86 J. Deke, “A Study of the Impact of Public School Pursuit of Educational Equity, ed. J. Yinger (MIT Press, Spending on Postsecondary Educational Attainment 2004), 87-111. This synopsis of Figlio’s main points Using Statewide School District Refinancing in actually comes from an earlier chapter in the same Kansas,” Economics of Education Review 22, no. 3 (2003): 275-284.

PAGE 32 87 T. A. Downes, J. Zabel and D. Ansel, Incomplete by one-fifth to one-quarter of a standard deviation” Does Money Grade: Massachusetts Education Reform at 15 (p. 1). Goertz and Weiss (2009) also evaluated the Matter in (Boston, MA: MassINC, 2009). effects of New Jersey school finance reforms but Education? did not attempt a specific empirical test of the 88 J. Guryan, “Does Money Matter? Estimates from relationship between funding level and distributional Education Finance Reform in Massachusetts,” changes and outcome changes. Thus, their findings Working Paper No. 8269 (Cambridge, MA: National are primarily descriptive. Goertz and Weiss explain Bureau of Economic Research, 2001). While this that on state assessments, achievement gaps closed paper remains an unpublished working paper, the substantially between 1999 and 2007, the period advantage of Guryan’s analysis is that he models over which Abbott funding was most significantly the expected changes in funding at the local level scaled up. A. M. Resch, Three Essays on Resources as a function of changes to the school finance in Education (dissertation) (Ann Arbor: University formula itself, through what is called an instrumental of Michigan, Department of Economics, 2008), variables or two-stage least squares approach. Then, retrieved Oct. 28, 2009, from http://deepblue.lib. Guryan evaluates the extent to which these policy- umich.edu/bitstream/2027.42/61592/1/aresch_1.pdf; induced variations in local funding are associated and M. Goertz and M. Weiss, Assessing Success in with changes in student outcomes. Across several School Finance Litigation: The Case of New Jersey model specifications, Guryan finds increased (New York City: Campaign for Educational Equity, outcomes for students at grade 4 but not at grade 8. Teachers College, Columbia University, 2009). The A counter study by the Beacon Hill Institute suggests authors explain: “State Assessments: In 1999 the gap that reduced class size and/or increased instructional between the Abbott districts and all other districts spending either has no effect on or actually worsens in the state was over 30 points. By 2007 the gap student outcomes. S. Jaggia and V. Vachharajani, was down to 19 points, a reduction of 11 points Money for Nothing: The Failures of Education Reform or 0.39 standard deviation units. The gap between in Massachusetts (2004), www.beaconhill.org/ the Abbott districts and the high-wealth districts fell BHIStudies/EdStudy5_2004/BHIEdStudy52004.pdf. from 35 to 22 points. Meanwhile performance in the low-, middle-, and high-wealth districts essentially 89 P. Nguyen-Hoang and J. Yinger,“Education Finance remained parallel during this eight-year period” Reform, Local Behavior, and Student Performance in (Figure 3, p. 23). Massachusetts,” Journal of Education Finance 39, no. 4 (2014): 297-322. 92 F. Neymotin, “The Relationship between School Funding and Student Achievement in Kansas Public 90 T. A. Downes, “School Finance Reform and School Schools,” Journal of Education Finance 36, no. 1 Quality: Lessons from Vermont,” in Helping Children (2010): 88-108; and J. P. Greene and J. R. Trivitt, Left Behind: State Aid and the Pursuit of Educational “Can Judges Improve Academic Achievement?,” Equity, ed. J. Yinger (Cambridge, MA: MIT Press, Peabody Journal of Education 83, no. 2 (2008): 224- 2004). 237. See our critique of these studies B. Baker and K. Welner, “School Finance and Courts: Does Reform 91 Two studies of school finance reforms in New Jersey Matter, and How Can We Tell,” Teachers College also merit some attention, in part because they Record 113, no. 11 (2011): 2374-2414.. directly refute findings of Hanushek and Lindseth and of the earlier Cato study, and do so with 93 W. Duncombe and J. Yinger, “Financing Higher more rigorous and detailed methods. The first, by Student Performance Standards: The Case of New Alex Resch of the University of Michigan (doctoral York State,” Economics of Education Review 19, no. dissertation in economics), explored in detail the 4 (2000): 363-386; A. Reschovsky and J. Imazeki, resource allocation changes during the scaling-up “Achieving Educational Adequacy through School period of school finance reform in New Jersey. Resch Finance Reform,” Journal of Education Finance found evidence suggesting that New Jersey Abbott (2001): 373-396; districts “directed the added resources largely to J. Imazeki and A. Reschovsky, “Is No Child Left instructional personnel” (p. 1), such as additional Behind an Un (or Under) Funded Federal Mandate? teachers and support staff. She also concluded that Evidence from Texas,” National Tax Journal (2004): this increase in funding and spending improved 571-588; J. Imazeki and A. Reschovsky, “Does No the achievement of students in the affected school Child Left Behind Place a Fiscal Burden on States? districts. Looking at the statewide 11th-grade Evidence from Texas,” Education Finance and assessment (“the only test that spans the policy Policy 1, no. 2 (2006): 217-246; and J. Imazeki and change”), she found “that the policy improves test A. Reschovsky, “Assessing the Use of Econometric scores for minority students in the affected districts Analysis in Estimating the Costs of Meeting State

PAGE 33 Does Money Education Accountability Standards: Lessons from 100 W. Duncombe and J. Yinger, “Are Education Cost Matter in Texas,” Peabody Journal of Education 80, no. 3 Functions Ready for Prime Time? An Examination Education? (2005): 96-125. of Their Validity and Reliability,” Peabody Journal of Education 86, no. 1 (2011): 28-57. See also 94 T. A. Downes and T. F. Pogue, “Adjusting School W. Duncombe and J. M. Yinger, “A Comment on Aid Formulas for the Higher Cost of Educating School District Level Production Functions Estimated Disadvantaged Students,” National Tax Journal Using Spending Data” (Maxwell School of Public (1994): 89-110; W. Duncombe and J. Yinger, Affairs, Syracuse University, 2007); and W. Duncombe “School Finance Reform: Aid Formulas and Equity and J. Yinger, “Making Do: State Constraints and Objectives,” National Tax Journal (1998): 239-262; Local Responses in California’s Education Finance W. Duncombe and J. Yinger, “Why Is It So Hard System,” International Tax and Public Finance 18, no. to Help Central City Schools?,” Journal of Policy 3 (2011): 337-368. For an alternative approach, see Analysis and Management 16, no. 1 (1997): 85-113; T. J. Gronberg, D. W. Jansen and L. L. Taylor, “The and W. Duncombe and J. Yinger, “How Much More Adequacy of Educational Cost Functions: Lessons Does a Disadvantaged Student Cost?,” Economics of from Texas,” Peabody Journal of Education 86, no. 1 Education Review 24, no. 5 (2005): 513-532. (2011): 3-27.

95 For a discussion, see B. D. Baker, “The Emerging 101 W. Duncombe and J. Yinger, “Are Education Cost Shape of Educational Adequacy: From Theoretical Functions Ready for Prime Time? An Examination Assumptions to Empirical Evidence,” Journal of of Their Validity and Reliability,” Peabody Journal Education Finance (2005): 259-287. See also M. of Education 86, no. 1 (2011): 28-57. See also Andrews, W. Duncombe and J. Yinger, “Revisiting W. Duncombe and J. M. Yinger, “A Comment on Economies of Size in American Education: Are School District Level Production Functions Estimated We Any Closer to a Consensus?,” Economics of Using Spending Data” (Maxwell School of Public Education Review 21, no. 3 (2002): 245-262; W. Affairs, Syracuse University, 2007). For an alternative Duncombe, J. Miner and J. Ruggiero, “Potential approach, see T. J. Gronberg, D. W. Jansen and L. L. Cost Savings from School District Consolidation: A Taylor, “The Adequacy of Educational Cost Functions: Case Study of New York,” Economics of Education Lessons from Texas,” Peabody Journal of Education Review 14, no. 3 (1995): 265-284; J. Imazeki and A. 86, no. 1 (2011): 3-27. Reschovsky, “Financing Adequate Education in Rural Settings,” Journal of Education Finance (2003): 137- 156; and T. J. Gronberg, D. W. Jansen and L. L. Taylor, “The Impact of Facilities on the Cost of Education,” 102 P. J. Wolf, A. Cheng, M. Batdorff, L. Maloney, J. F. National Tax Journal 64, no. 1 (2011): 193-218. May and S. T. Speakman, The Productivity of Public Charter Schools (Fayetteville, AR: Department 96 T. Downes, What Is Adequate? Operationalizing the of Education Reform, University of Arkansas, 2014), Concept of Adequacy for New York State (2004), retrieved Aug. 5, 2014, from www.uaedreform. http://www.albany.edu/edfin/Downes%20 org/downloads/2014/07/the-productivity-of- EFRC%20Symp%2004%20Single.pdf. public-charter-schools.pdf.

97 See, for example, B. D. Baker, “Exploring the 103 Some, however, find that charter school expansion Sensitivity of Education Costs to Racial Composition may exert negative effects through student/peer of Schools and Race-Neutral Alternative Measures: sorting: A Cost Function Application to Missouri,” Peabody “Results show that charters induce modest but Journal of Education 86, no. 1 (2011): 58-83. statistically significant drops in math and language test scores, particularly for elementary students. 98 Completed and released in 2006, eventually However, results for middle and high school students published as J. Imazeki, “Assessing the Costs of show improvements in discipline.” S. A. Imberman, Adequacy in California Public Schools: A Cost “The Effect of Charter Schools on Achievement and Function Approach,” Education 3, no. 1 (2008): Behavior of Public School Students,” Journal of 90-108. Public Economics 95, no. 7 (2011): 850-863. Epple, Romano and Zimmer (2015) discuss the much larger 99 See the acknowledgements at http://files.eric.ed.gov/ literature on this topic: D. Epple, R. Romano and R. fulltext/ED508961.pdf. Final published version: R. Zimmer, Charter Schools: A Survey of Research on Costrell, E. Hanushek and S. Loeb, “What Do Cost Their Characteristics and Effectiveness (no. w21256) Functions Tell Us about the Cost of an Adequate (National Bureau of Economic Research, 2015). Education?,” Peabody Journal of Education 83, no. 2 (2008): 198-223.

PAGE 34 104 D. Epple, R. Romano and R. Zimmer, Charter Schools: A Cost Frontier Approach,” Economics of Education Does Money A Survey of Research on Their Characteristics and Review 31, no. 2 (2012): 302-317. Matter in Effectiveness (no. w21256) (National Bureau of Education? Economic Research, 2015). More recent studies 110 W. Dobbie and R. G. Fryer Jr., Getting Beneath the have affirmed the weak outcomes of Online Veil of Effective Schools: Evidence from New York Charter schools: Center for Research on Education City (no. w17632) (National Bureau of Economic Outcomes (2015) Online Charter School Study. Research, 2011). The authors note: “We find that https://credo.stanford.edu/pdfs/Online%20 traditionally collected input measures—class size, per Charter%20Study%20Final.pdf Gill, B., Walsh, L., pupil expenditure, the fraction of teachers with no Wulsin, C. S., Matulewicz, H., Severn, V., Grau, E., certification, and the fraction of teachers with an & Kerwin, T. (2015). Inside Online Charter Schools advanced degree—are not correlated with school (No. 9c4fae4409c44298a559a14ea948fb2c). effectiveness.” Mathematica Policy Research. 111 B. D. Baker, K. Libby and K. Wiley, Spending by 105 P. J. Wolf, A. Cheng, M. Batdorff, L. Maloney, J. F. the Major Charter Management Organizations: May and S. T. Speakman, The Productivity of Public Comparing Charter School and Local Public District Charter Schools (Fayetteville, AR: Department of Financial Resources in New York, Ohio, and Texas Education Reform, University of Arkansas, 2014), (National Education Policy Center, 2012); and B. retrieved Aug. 5, 2014, from www.uaedreform. D. Baker, K. Libby and K. Wiley, “Charter School org/downloads/2014/07/the-productivity-of- Expansion and Within-District Equity: Confluence public-charter-schools.pdf; and G. V. Glass, Review or Conflict?,” Education Finance and Policy (2015). of “The Productivity of Public Charter Schools” Others report similar resource differentials (and more) (Boulder, CO: National Education Policy Center, 2014), for KIPP schools: T. J. Gronberg, D. W. Jansen and L. http://nepc.colorado.edu/thinktank/review- L. Taylor, “The Relative Efficiency of Charter Schools: productivity-public-charter. A Cost Frontier Approach,” Economics of Education Review 31, no. 2 (2012): 302-317. 106 B. D. Baker, Review of “Charter Funding: Inequity Expands” (Boulder, CO: National Education 112 Others also explain that many charter schools may Policy Center, 2014), http://nepc.colorado.edu/ achieve positive outcome effects by increasing the thinktank/review-charter-funding-inequity. length of school days, providing additional weekend instruction and extending school years into the summer months. The additional costs associated with all of this additional time are kept in check by 107 D. Epple, R. Romano and R. Zimmer, Charter Schools: employing inexperienced teachers and leveraging A Survey of Research on Their Characteristics and the lack of collective bargaining (in many charters) Effectiveness (no. w21256) (National Bureau of to require the longer hours and years, with less Economic Research, 2015). regard for “burnout” or turnover because turnover keeps staffing costs low. Again, scalability of 108 C. C. Tuttle, B. Gill, P. Gleason, V. Knechtel, I. Nichols- these strategies is questionable. In all likelihood, Barrer and A. Resch, KIPP Middle Schools: Impacts systemwide expansion of school days and school on Achievement and Other Outcomes, Final Report years would require additional resources. (Mathematica Policy Research, 2013). Abdulkadiroğlu, A., Angrist, J. D., Dynarski, S. M., Angrist, J. D., Dynarski, S. M., Kane, T. J., Pathak, Kane, T. J., & Pathak, P. A. (2011). Accountability P. A., & Walters, C. R. (2010). Inputs and impacts and flexibility in public schools: Evidence from in charter schools: KIPP Lynn. American Economic Boston’s charters and pilots. The Quarterly Journal of Review, 100(2), 239-243. Economics, 126(2), 699-748. Hoxby, C. M., Murarka, S., & Kang, J. (2009). How 109 B. D. Baker, K. Libby and K. Wiley, Spending by New York City’s charter schools affect achievement. the Major Charter Management Organizations: Cambridge, MA: New York City Charter Schools Comparing Charter School and Local Public District Evaluation Project, 1-85. Financial Resources in New York, Ohio, and Texas (National Education Policy Center, 2012); and B. 113 T. J. Gronberg, D. W. Jansen and L. L. Taylor, “The D. Baker, K. Libby and K. Wiley, “Charter School Relative Efficiency of Charter Schools: A Cost Frontier Expansion and Within-District Equity: Confluence Approach,” Economics of Education Review 31, no. 2 or Conflict?,” Education Finance and Policy (2015). (2012): 302-317. Others report similar resource differentials (and more) for KIPP schools: T. J. Gronberg, D. W. Jansen and L. 114 T. J. Gronberg, D. W. Jansen and L. L. Taylor, “The L. Taylor, “The Relative Efficiency of Charter Schools:

PAGE 35 Does Money Relative Efficiency of Charter Schools: A Cost Frontier 119 http://www.edlawcenter.org/assets/files/ Matter in Approach,” Economics of Education Review 31, no. 2 pdfs/maisto/masito%20trial%20documents/ Education? (2012): 302-317. State%27s%20Expert%20Report%20-%20 Dr.%20Eric%20Hanushek.pdf. 115 L. L. Taylor, B. L. Alford, K. G. Rollins, D. B. Brown, J. R. Stillisano and H. C. Waxman, “Evaluation of Texas 120 B. D. Baker, Revisiting the Age-Old Question: Does Charter Schools 2009-2010” (Revised Draft) (Texas Money Matter in Education? (Albert Shanker Institute, Education Research Center, Texas A&M University, 2012). College Station, 2011). 121 Specifically, Hanshek includes the following footnote: 116 D. Epple, R. Romano and R. Zimmer, Charter Schools: “Hanushek (2003). See also Hanushek (1981, A Survey of Research on Their Characteristics and (1986, (1989). The statistical analyses focus on the Effectiveness (no. w21256) (National Bureau of independent impact of resources on performance Economic Research, 2015). after allowing for differences among families, peers, and neighborhoods. A variety of sophisticated 117 E. A. Hanushek, “Economics of Schooling: Production approaches have been applied to schooling situations and Efficiency in Public Schools,” Journal of Economic across the countries, and the reviews summarize Literature 24, no. 3 (1986): 1141-1177. A few these studies. The aggregate results of the most years later, Hanushek paraphrased this conclusion sophisticated of these studies are shown below.” in another widely cited article as “Variations in http://www.edlawcenter.org/assets/files/ school expenditures are not systematically related to pdfs/maisto/masito%20trial%20documents/ variations in student performance.” E. A. Hanushek, State%27s%20Expert%20Report%20-%20 “The Impact of Differential Expenditures on School Dr.%20Eric%20Hanushek.pdf. Performance,” Educational Researcher 18, no. 4 (1989): 45-62. 122 C. K. Jackson, R. C. Johnson and C. Persico, The Effects of School Spending on Educational and 118 Kevin Welner and I discuss at length Hanushek Economic Outcomes: Evidence from School Finance and Lindseth’s strange heavy reliance on these two Reforms (no. w20847) (National Bureau of Economic Cato reports from the late 1990s on Kansas City Research, 2015). and New Jersey, as if these reports are the seminal works of the field. Yet, amazingly, Hanushek and 123 http://hanushek.stanford.edu/opinions/does-money- Lindseth ignore outright most of the major peer- matter-after-all. reviewed articles on school finance reform by credible researchers using credible methods, addressed herein 124 http://hanushek.stanford.edu/sites/default/ and addressed in my article with Welner: B. Baker files/publications/Hanushek%202003%20EJ%20 and K. Welner, “School Finance and Courts: Does 113%28485%29.pdf. Reform Matter, and How Can We Tell,” Teachers College Record 113, no. 11 (2011): 2374-2414. 125 http://www.edlawcenter.org/assets/files/ For example, in a Hoover Institution commentary pdfs/maisto/masito%20trial%20documents/ regarding school funding litigation in New York state, State%27s%20Expert%20Report%20-%20 Hanushek (2002) noted: “One need only look at the Dr.%20Eric%20Hanushek.pdf. results in Kansas City. A school desegregation ruling 126 www.edlawcenter.org/assets/files/pdfs/maisto/ in the 1980s began a period of more than a decade masito%20trial%20documents/State%27s%20 when the schools had access to virtually unlimited Expert%20Report%20-%20Dr.%20Eric%20 state funds. The dreams of school personnel did Hanushek.pdf. not translate into any measurable gains in student performance, even as their schools moved to the 127 http://educationnext.org/money-matter. very top of national spending.” A PDF of this op-ed is available on Eric Hanushek’s website, at http:// 128 http://www.shawneecourt.org/ edpro.stanford.edu/hanushek/admin/pages/ DocumentCenter/View/457. files/uploads/hanushek_0302.pdf. Baker and Green explain the various mythologies embedded in 129 B. Baker and K. Welner, “School Finance and this particular statement. See P. C. Green and B. D. Courts: Does Reform Matter, and How Can We Tell,” Baker, “Urban Legends, Desegregation and School Teachers College Record 113, no. 11 (2011): 2374- Finance: Did Kansas City Really Prove That Money 2414. Doesn’t Matter?,” Michigan Journal of Race and Law 12, no. 1 (2006): 57-105. 130 http://www.edlawcenter.org/assets/files/

PAGE 36 pdfs/maisto/masito%20trial%20documents/ 139 http://jaypgreene.com/2015/05/29/does-school- Does Money State’s%20Expert%20Report%20-%20Dr.%20 spending-matter-after-all. Matter in Eric%20Hanushek.pdf. Education? 140 http://www.edlawcenter.org/assets/files/ 131 http://junkcharts.typepad.com/junk_ pdfs/maisto/masito%20trial%20documents/ charts/2011/04/bill-gates-should-hire-a- State’s%20Expert%20Report%20-%20Dr.%20 statistical-advisor.html. David%20Armor.pdf.

132 See, for example, http://www.epi.org/ 141 R. Murnane, “Interpreting the Evidence on Does publication/fact-challenged_policy. Money Matter?,” Harvard Journal of Legislation 28 (1991): 457-464. 133 The authors explain: “To see the problems of Hanushek’s logic, consider the following true statistics: 142 B. Baker and K. Welner, “School Finance and Courts: between 1960 and 2000 the rate of cigarette smoking Does Reform Matter, and How Can We Tell,” Teachers for females decreased by more than 30 percent while College Record 113, no. 11 (2011): 2374-2414. the rate of deaths by lung cancer increased by more than 50 percent over the same time period. [1] An 143 http://www.shawneecourt.org/ analysis of these time trends might lead one to infer DocumentCenter/View/457. that smoking reduces lung cancer. However, most informed readers can point out numerous flaws in 144 H. Rose and J. R. Betts, “The Effect of High School looking at this time trend evidence and concluding Courses on Earnings,” Review of Economics and that ‘if smoking causes lung cancer, then there should Statistics 86, no. 2 (2004): 510; A. Gamoran and have been a large corresponding reduction in cancer E. C. Hannigan, “Algebra for Everyone? Benefits rates so that there can be no link between smoking of College-Preparatory Mathematics for Students and lung cancer.’ However, this is exactly the facile with Diverse Abilities in Early Secondary School,” logic invoked by Hanushek regarding the effect of Educational Evaluation and Policy Analysis 22, no. 3 school spending on student achievement” (Jackson, K., (Fall 2000): 241-254; M. C. Long, P. Iatarola and D. Johnson, R.C., Persico, C. (2015) Money Does Matter Conger, “Explaining Gaps in Readiness for College- After All. Education Next. http://educationnext.org/ Level Math: The Role of High School Courses,” money-matter) . Education Finance and Policy 4, no. 1 (Winter 2009): 1-33; and Leslie Killgore, “Merit and Competition 134 B. Baker and K. Welner, “School Finance and Courts: in Selective College Admissions,” Review of Higher Does Reform Matter, and How Can We Tell,” Teachers Education 32, no. 4 (Summer 2009): 471. College Record 113, no. 11 (2011): 2374-2414. 145 B. D. Baker, “Cheerleading, Ceramics and the 135 http://www.robblaw.com/PDFs/1169.pdf. Non-Productive Use of Educational Resources in High Need Districts: Really?” (paper presented at 136 http://www.edlawcenter.org/assets/files/ the Annual Meeting of the American Educational pdfs/maisto/masito%20trial%20documents/ Research Association, New Orleans, LA, 2011), State’s%20Expert%20Report%20-%20Dr.%20 http://schoolfinance101.files.wordpress. Eric%20Hanushek.pdf. com/2010/01/b-baker-mo_il-resourcealloc- aera2011.pdf. 137 As explained by the authors: “We also examine changes in student performance in 41 states within 146 C. Taylor, “Does Money Matter? An Empirical Study the United States between 1992 and 2011, allowing Introducing Resource Costs and Student Needs us to compare these states with each other. Our into Educational Production Function Analysis,” in findings come from assessments of performance in Developments in School Finance, 1997, ed. W. J. math, science, and reading of representative samples Fowler Jr. (Washington, DC: U.S. Department of in particular political jurisdictions of students who at Education, National Center for Education Statistics, the time of testing were in 4th or 8th grade or were 1998), http://nces.ed.gov/pubs98/98212. roughly ages 9-10 or 14-15.” pdf#page=83.

138 E. A. Hanushek, P. E. Peterson and L. Woessmann, “Is 147 H. Wenglinsky, “How Money Matters: The Effect of the US Catching Up: International and State Trends School District Spending on Academic Achievement,” in Student Achievement,” Education Next 12, no. 4 Sociology of Education 70, no. 3 (1997): 221-237. (2012): 24, http://www.hks.harvard.edu/pepg/ PDF/Papers/PEPG12-03_CatchingUp.pdf. 148 G. Borman and G. Hewes, “The Long-Term Effects and Cost-Effectiveness of Success for All,”

PAGE 37 Does Money Educational Evaluation and Policy Analysis 24, no. 4 588; and T. Downes and T. Pogue, “Adjusting School Matter in (2002): 243-266. Aid Formulas for the Higher Cost of Educating Education? Disadvantaged Students,” National Tax Journal 47 149 R. Fryer, “Creating ‘No Excuses’ (Traditional) Public (1994): 89-110. Schools: Preliminary Evidence from an Experiment in Houston,” NBER Working Paper 17494 (Cambridge, 153 These studies have some significant empirical MA: National Bureau of Economic Research, 2011), advantages over production function studies in www.nber.org/papers/w17494. that they allow for corrections to be made for differences in the level of efficiency in producing 150 E. A. Hanushek, “Empirical and Conceptual Issues in outcomes across districts. This is possible in a the Estimation of Educational Production Functions,” cost-function framework because the spending Journal of Human Resources 14, no. 3 (1979): 351- variable is the dependent variable rather than one 388. of the independent variables. When spending is the dependent variable, one can include in the 151 B. D. Baker, “Can Flexible Non-Linear Modeling Tell model characteristics of school districts theoretically Us Anything New about Educational Productivity?,” assumed to be related to greater inefficiency, such as Economics of Education Review 20, no. 1 (2001): less constrained fiscal capacity. 81-92; and D. N. Figlio, “Functional Form and the Estimated Effects of School Resources,” Economics of 154 See, for example, G. D. Borman, G. Hewes, L. T. Education Review 18, no. 2 (1999): 242-252. Overman and S. Brown, “Comprehensive School Reform and Achievement: A Meta-Analysis,” Review 152 B. D. Baker, “Exploring the Sensitivity of Education of Educational Research 73, no. 2 (2003): 125-230; Costs to Racial Composition of Missouri School and G. Borman and G. Hewes, “The Long-Term Districts,” Peabody Journal of Education 86, no. 1 Effects and Cost-Effectiveness of Success for All,” (2011): 58-83; W. Duncombe and J. Yinger, “How Educational Evaluation and Policy Analysis 24, no. 4 Much More Does a Disadvantaged Student Cost?,” (2002): 243-266. Economics of Education Review 24, no. 5 (2005): 513-532; W. Duncombe and J. M. Yinger, “Financing Higher Performance Standards: The Case of New York State,” Economics of Education Review 19, no. 3 (2000): 363-386; W. Duncombe, A. Lukemeyer and J. Yinger, “The No Child Left Behind Act: Have Federal Funds Been Left Behind?,” Public Finance Review 36, no. 4 (2008): 381-407; W. Duncombe and J. M. Yinger, “Performance Standards and Education Cost Indexes: You Can’t Have One Without the Other,” in Equity and Adequacy in Education Finance: Issues and Perspectives, ed. H. F. Ladd, R. Chalk and J. S. Hansen (Washington, DC: National Academy Press, 1999), 260-297; W. Duncombe and J. M. Yinger, “School Finance Reforms: Aid Formulas and Equity Objectives,” National Tax Journal 51, no. 2 (1998): 239-263; W. Duncombe and J. M. Yinger, “Why Is It So Hard to Help Central City Schools?,” Journal of Policy Analysis and Management 16, no. 1 (1997): 85-113; T. Gronberg, D. Jansen, L. Taylor and K. Booker, School Outcomes and Schools Costs: The Cost Function Approach (College Station, TX: Busch School of Government and Public Service, Texas A&M University, 2004), retrieved March 1, 2006, from http://bush.tamu.edu/research/ faculty_projects/txschoolfinance/papers/ SchoolOutcomesAndSchoolCosts.pdf; J. Imazeki and A. Reschovsky, “Is No Child Left Beyond an Un (or Under) Funded Federal Mandate? Evidence from Texas,” National Tax Journal 57, no. 3 (2004): 571-

PAGE 38 About the Author Does Money Matter in Education?

Bruce Baker is a professor in the Graduate School of Education at Rutgers in New Jersey. From 1997 to 2008, he was a professor at the University of Kansas in Lawrence. He co-authored Financing Education Systems, a graduate-level textbook on school finance policy, and has written a multitude of peer-reviewed research articles on state school finance policy, teacher labor markets, school leadership labor markets, and higher education finance and policy. His recent work has focused on measuring cost variations associated with schooling contexts and student population characteristics, including ways to improve state school finance policies and local district allocation formulas (including weighted student funding) to better meet the needs of students. Baker has also consulted for state legislatures, boards of education, and other organizations on education policy and school finance issues, and has testified in state school finance litigation in Arizona, Kansas and Missouri.

PAGE 39 555 New Jersey Ave., NW | Washington, D.C. 20001 Tel: 202.879.4401 | shankerinstitute.org